Skip to content

armgit / hypit

Public Imported from GitHub
0

Clone any viral video with AI agents. Not just a script, the whole workflow: swap the face, the words, the B-roll, ship 100 variants in one command, and get your 100M views.

docs: remove the published specifications

rpone00ac2b00b62f
diff --git a/README.md b/README.md
index 0b404281..474b82f4 100644
--- a/README.md
+++ b/README.md
@@ -141,15 +141,12 @@ expresses the work, the Run Graph selects Targets and Candidates for one Build.
verifies and advances the resulting state machine, so video concepts stay in independently
installable packages — Core hard-codes no model, Track or Provider.
-The normative laws are in the
-[Core Kernel specification](https://narratage.hypit.ai/guide/spec/core-kernel).
-
## Where to go next
- [Quickstart](https://narratage.hypit.ai/quickstart) — the files you control, the commands, and
your first real Build.
- [Develop](https://narratage.hypit.ai/guide/develop) — package architecture, adding an author
- package or a Provider, and the normative specifications.
+ package or a Provider.
## License
diff --git a/README.zh-CN.md b/README.zh-CN.md
index 6c03b092..effd4e71 100644
--- a/README.zh-CN.md
+++ b/README.zh-CN.md
@@ -124,12 +124,10 @@ node --run narratage -- plan examples/talking-film-graph-check/build.svrun \
AI 生成慢、贵、不确定,而且每一步的输出往往就是下一步的输入。Narratage 把由此产生的选择呈现为两张平级的图:Author Graph 表达工作本身,Run Graph 为一次 Build 选择 Target 与 Candidate。Core 只负责解析、校验并推进由此得到的状态机,因此视频领域的概念都留在可独立安装的包里——Core 不硬编码任何模型、Track 或 Provider。
-规范性法则见 [Core Kernel 规范](https://narratage.hypit.ai/zh/guide/spec/core-kernel)。
-
## 接下来去哪
- [快速开始](https://narratage.hypit.ai/zh/quickstart) —— 你掌控的文件、命令,以及第一次真实 Build。
-- [开发](https://narratage.hypit.ai/zh/guide/develop) —— 包架构、添加 Author 包或 Provider,以及全部规范文档。
+- [开发](https://narratage.hypit.ai/zh/guide/develop) —— 包架构、添加 Author 包或 Provider。
## 许可
diff --git a/docs/.vitepress/config.ts b/docs/.vitepress/config.ts
index 92b1c333..e3022442 100644
--- a/docs/.vitepress/config.ts
+++ b/docs/.vitepress/config.ts
@@ -161,28 +161,6 @@ const enTheme = {
{ text: "Conventions", link: "/guide/conventions" },
],
},
- {
- text: "Specifications",
- collapsed: true,
- items: [
- { text: "Core Kernel", link: "/guide/spec/core-kernel" },
- { text: "Author Surface Binding", link: "/guide/spec/author-surface-binding" },
- { text: "Script Surface", link: "/guide/spec/script-surface" },
- { text: "Shared Spatial Layout", link: "/guide/spec/spatial-layout" },
- { text: "Track Authoring Model", link: "/guide/spec/track-authoring" },
- { text: "Track and Composition", link: "/guide/spec/track-composition" },
- { text: "Visual IR", link: "/guide/spec/visual-ir" },
- { text: "Track Expressiveness Gate", link: "/guide/spec/track-expressiveness" },
- { text: "Caption Program", link: "/guide/spec/caption-program" },
- { text: "Fine Caption Style", link: "/guide/spec/caption-fine" },
- { text: "Media Track", link: "/guide/spec/media-track" },
- { text: "Typography Track", link: "/guide/spec/typography-track" },
- { text: "Audio Track", link: "/guide/spec/audio-track" },
- { text: "Deck Track", link: "/guide/spec/deck-track" },
- { text: "Ranking Track", link: "/guide/spec/ranking-track" },
- { text: "Screen Overlay", link: "/guide/spec/screen-overlay" },
- ],
- },
],
},
outline: { level: [2, 3] as [number, number], label: "On this page" },
@@ -226,28 +204,6 @@ const zhTheme = {
{ text: "代码规范", link: "/zh/guide/conventions" },
],
},
- {
- text: "规范",
- collapsed: true,
- items: [
- { text: "Core Kernel", link: "/zh/guide/spec/core-kernel" },
- { text: "Author Surface 绑定", link: "/zh/guide/spec/author-surface-binding" },
- { text: "Script Surface", link: "/zh/guide/spec/script-surface" },
- { text: "共享空间布局", link: "/zh/guide/spec/spatial-layout" },
- { text: "Track 创作模型", link: "/zh/guide/spec/track-authoring" },
- { text: "Track 与 Composition", link: "/zh/guide/spec/track-composition" },
- { text: "Visual IR", link: "/zh/guide/spec/visual-ir" },
- { text: "Track 表现力门槛", link: "/zh/guide/spec/track-expressiveness" },
- { text: "Caption Program", link: "/zh/guide/spec/caption-program" },
- { text: "Fine Caption 样式族", link: "/zh/guide/spec/caption-fine" },
- { text: "Media Track", link: "/zh/guide/spec/media-track" },
- { text: "Typography Track", link: "/zh/guide/spec/typography-track" },
- { text: "Audio Track", link: "/zh/guide/spec/audio-track" },
- { text: "Deck Track", link: "/zh/guide/spec/deck-track" },
- { text: "Ranking Track", link: "/zh/guide/spec/ranking-track" },
- { text: "Screen Overlay", link: "/zh/guide/spec/screen-overlay" },
- ],
- },
],
},
outline: { level: [2, 3] as [number, number], label: "本页目录" },
diff --git a/docs/guide/spec/audio-track.md b/docs/guide/spec/audio-track.md
deleted file mode 100644
index 10caacec..00000000
--- a/docs/guide/spec/audio-track.md
+++ /dev/null
@@ -1,256 +0,0 @@
----
-title: SVML Audio Track Authoring
-description: implemented pre-release contract for the generic official Audio Track package, including.
----
-
-# SVML Audio Track Authoring
-
-Status: implemented pre-release contract for the generic official Audio Track package, including
-its self-described author Surface, exact sample-domain lowering and local/remote execution witness.
-It is not yet a frozen public ABI.
-
-## Purpose
-
-`@narratage/audio-track` lets an author place already selected and normalized audio material into
-ProgramSpace. Music, sound effects, additional voice-over and source audio are ordinary peers. The
-package owns no privileged base lane and Composition mixes its result through the existing
-[`AudioTrack`](./track-composition.md) terminal contract.
-
-This is a video-domain author package, not a Core primitive, Runtime service or Provider. The
-separation is:
-
-```text
-explicit media source ─> inspect/select/normalize ─> SynchronizedMedia
- │
-Narrative points / ProgramSpace ─> temporal window ─────┼─> Audio Track Program
-source trim ─> occupancy ─> level/fades ────────────────┘ │
- ▼
- AudioTrack
- │
- ▼
- audio plan -> Provider
-```
-
-Core sees fixed graph inputs and one output Record. It does not know clips, buses, music, speech,
-looping, fading or mixing.
-
-## 1. Input truth
-
-Every audio item receives its material through an explicit graph edge. A first implementation
-should consume `SynchronizedMedia` from `@narratage/media` and require its normalized audio member:
-
-- 48 kHz;
-- stereo;
-- signed 16-bit PCM WAV;
-- exact sample-frame count;
-- preserved source level.
-
-An MP4 merely containing AAC is not an audio item. The author graph must explicitly inspect the
-container, select an audio stream and normalize it. Likewise, a generated speech file does not
-become `SpeechAudioBasis` merely because it contains a voice.
-
-The Audio Track lowerer may derive the terminal Artifact duration from the normalized sample-frame
-count. It must not copy source identities, Narrative digests, provider names or other lineage into
-the clip. Those relations remain graph edges and Derivations.
-
-Do not introduce a new public canonical-audio Type merely to shorten one edge. If several unrelated
-packages later need a first-class canonical-audio value, that focused intrinsic Type belongs in
-`@narratage/media`; it still does not belong in Core or an umbrella contracts package.
-
-## 2. Author Program
-
-The package owns an arbitrary-length collection of independent items. Conceptually each item has:
-
-```ts
-type AudioItemProgram = {
- readonly id: string;
- readonly source: SynchronizedMediaInput;
- readonly temporal: TemporalBinding;
- readonly trim?: AudioSourceTrim;
- readonly occupancy: AudioOccupancy;
- readonly mix: AudioItemMix;
-};
-```
-
-These are package-level concerns. They are not new fields on Core Operations or the terminal
-`AudioTrack` contract.
-
-An illustrative author Surface may remain compact:
-
-```xml
-<audio:Track id="mix" space={speech.space}>
- <audio:Clip
- source={music.media}
- during={program}
- playback="loop"
- gain={0.28}
- fade-in="800ms"
- fade-out="1200ms"
- />
-
- <audio:Clip
- source={impact.media}
- at={story.moment.reveal}
- for="420ms"
- playback="once"
- />
-</audio:Track>
-```
-
-This spelling is illustrative rather than a frozen Surface. `during`, `at` and `for` must lower to
-the shared point-expression and window-projection algebra; they cannot create a second timing
-engine.
-
-## 3. Temporal placement
-
-Audio uses the same rules as [`track-authoring.md`](./track-authoring.md):
-
-1. locate Selection, Moment or Program points;
-2. expand `one` or `each` occurrences;
-3. project a directed candidate window;
-4. intersect and validate it against ProgramSpace;
-5. apply audio source trim and occupancy;
-6. lower both target placement and source sampling into the exact 48 kHz sample domain.
-
-Audio items are independent. Overlapping projected windows mix; they are never clipped according
-to authoring order, auto-stitched, deduplicated or treated as competing alternatives. Several Audio
-Track values behave the same as several overlapping items in one Audio Track.
-
-ProgramSpace frame boundaries are converted to 48 kHz sample boundaries by the one deterministic
-rule already owned by the media pipeline. The author package must not implement another seconds-to-
-samples rounding path.
-
-## 4. Source trim
-
-Source trim defines the effective intrinsic interval before occupancy. It is an explicit head and
-optional tail within the exact normalized sample domain. Author-facing seconds or frames are
-quantized once into sample boundaries.
-
-The lowerer rejects:
-
-- a negative or out-of-source boundary;
-- an end not greater than the start;
-- an empty effective interval;
-- an unavailable audio member;
-- a second, package-private interpretation of source duration.
-
-Source trim is not target timing. Moving a target window never silently changes which source region
-was authored, except where the selected occupancy policy explicitly aligns or retimes it.
-
-## 5. Audio occupancy
-
-Audio has its own truthful occupancy vocabulary. It must not reuse the visual fiction of holding one
-sample forever.
-
-```ts
-type AudioOccupancy =
- | { readonly mode: "once"; readonly align: "start" | "end" }
- | { readonly mode: "loop"; readonly align: "start" | "end" }
- | {
- readonly mode: "stretch";
- readonly minRate: number;
- readonly maxRate: number;
- readonly pitch: "preserve";
- };
-```
-
-The laws are:
-
-| Policy | Effective source shorter than window | Effective source longer than window |
-|---|---|---|
-| `once/start` | play at the start, then silence | play the source head and truncate at the window end |
-| `once/end` | silence first, then play to the end | play the source tail and end at the window end |
-| `loop/start` | repeat from the source-head phase | play the source head and truncate |
-| `loop/end` | choose phase so the source tail meets the window end | play the source tail |
-| `stretch` | slow into the complete window | accelerate into the complete window |
-
-`once` replaces the old ambiguous `native` / `finish` behavior. End alignment is not reverse
-playback. `loop/end` changes loop phase; it does not reverse samples. `stretch` is pitch-preserving,
-must stay inside the authored rate bounds and fails rather than silently applying a faster rate and
-then truncating. The old `fit_base` rule—speed up by at most 1.1x and silently cut the rest—is
-retired.
-
-Explicit silence is represented by the absence of a clip over that interval. Padding a file with
-silence is a separate media transformation when the padded bytes themselves matter.
-
-## 6. Level and fades
-
-The first author contract needs only three orthogonal controls:
-
-- non-negative linear `gain`;
-- `fadeIn` over the audible rendered interval;
-- `fadeOut` over the audible rendered interval.
-
-Zero gain is explicit silence. A UI or recipe may expose decibels, but conversion to terminal linear
-gain is deterministic and package-owned. Fade durations must be non-negative, cannot exceed the
-audible interval and cannot secretly expand the target window.
-
-There is no implicit loudness normalization, limiter, compressor, sidechain ducking, crossfade or
-automatic music-bed behavior. Those are legitimate future audio-processing or mix Programs, but
-they must consume explicit inputs and make their behavior author-visible. A runtime or Provider may
-not add them as an environment preference.
-
-The former terminal `bus: speech | music | sfx | source` label has been removed. It had no rendering
-effect and therefore carried no honest meaning. A future mix-routing contract must have an explicit
-consumer and graph behavior; merely naming a clip `music` cannot cause ducking.
-
-## 7. Lowering and execution
-
-The deterministic author package lowers every realized item to an ordinary terminal `AudioClip`:
-
-- exact target `{ startSample, endSampleExclusive }` from temporal projection;
-- canonical WAV Artifact from the normalized input;
-- exact source sample interval, loop phase and playback rate from trim/occupancy;
-- pitch-preservation intent, linear gain and exact sample fades from presentation.
-
-Composition validates the Audio Track against an explicitly connected ProgramSpace. The media
-pipeline then compiles all peer Audio Tracks into one content-addressed `AudioProgramPlan`. Local
-FFmpeg, Lambda or another Provider executes that same plan; Provider placement cannot reinterpret
-the mix.
-
-The package emits one `AudioTrack` even when it owns many clips. That is an author-component output,
-not a privileged lane. A package that naturally owns one visual hit and one sound hit may emit peer
-`VisualTrack` and `AudioTrack` outputs from one Fragment, while Film receives both through ordinary
-edges.
-
-## 8. Legacy migration audit
-
-Retain:
-
-- arbitrary audio inputs;
-- Program-full, Selection, Moment and absolute placement;
-- overlapping music/SFX/voice contributions;
-- per-item gain and fades;
-- one-shot, loop and bounded pitch-preserving stretch;
-- explicit source trim and start/end alignment.
-
-Retire:
-
-- dynamic numbered ports as the saved author truth;
-- `z_index` on audio;
-- implicit full timing hidden in a node executor;
-- config-row priority clipping;
-- `fit_base` and its silent speed-then-truncate behavior;
-- `fill: freeze` for audio;
-- automatic extraction of whatever audio happens to be inside a video;
-- inert bus labels and hidden mastering.
-
-## 9. Implementation and acceptance
-
-The shared Temporal package and the pre-freeze terminal Audio candidate have now been completed
-without a Core branch, Provider-family branch or privileged Film lane:
-
-1. **Implemented:** add the `@narratage/audio-track` author package and self-described Surface;
-2. **Implemented:** consume explicit normalized `SynchronizedMedia` inputs;
-3. **Implemented:** sample-exact trim and every occupancy law;
-4. **Implemented:** lower to the peer terminal `AudioTrack`;
-5. **Implemented:** graph tests for arbitrary item counts and Selection/Moment `one` / `each`;
-6. **Implemented:** sample-level tests for shorter/equal/longer sources, both alignments, loops, stretch bounds
- and fades;
-7. **Implemented:** two overlapping items and two peer Audio Tracks produce the same planned mix facts;
-8. **Implemented:** local FFmpeg and remote Lambda Providers receive the identical
- content-addressed `AudioProgramPlan`; the real FFmpeg witness additionally checks exact loop
- phase sample by sample.
-
-The migration is incomplete if adding this package requires a Core branch, a Film audio family, a
-new queue, a Provider name in SVML or a hidden Base-audio rule.
diff --git a/docs/guide/spec/author-surface-binding.md b/docs/guide/spec/author-surface-binding.md
deleted file mode 100644
index 5c99f92a..00000000
--- a/docs/guide/spec/author-surface-binding.md
+++ /dev/null
@@ -1,154 +0,0 @@
----
-title: Author Surface Binding
-description: implemented compiler contract for trusted package-owned Surfaces; third-party isolation is.
----
-
-# Author Surface Binding
-
-Status: implemented compiler contract for trusted package-owned Surfaces; third-party isolation is
-not yet implemented.
-
-## 1. Purpose
-
-The generic Markup Frontend owns element syntax. A package-owned Surface owns what one imported tag
-means. SVS owns only its stylesheet syntax and produces generic immutable Recipe records. Neither
-Markup, SVS nor Core owns Caption, Media, Film or another package's property vocabulary.
-
-The bridge is compile-time reference inspection: a Surface such as Caption must be able to
-read an explicitly referenced public Recipe and lower it into a nominal `CaptionProgram` before the
-Core graph is built.
-
-This document defines that bridge without adding a global Recipe registry or a Runtime step.
-
-## 2. Ordered compilation boundary
-
-One Source Closure compiles in this order:
-
-```text
-recursive imported SourceUnits
- -> admitted public Records and symbolic component exports
- -> current document generic syntax
- -> package-owned Surface compilation
- -> typed authored Records + AuthorComponents + locked GraphFragments
- -> parser-independent Author linking
- -> Core CompiledGraph
-```
-
-An imported `.svs` SourceUnit therefore finishes before a consuming `.svml` SourceUnit. The
-consuming Surface receives the structured element and a resolver for references explicitly written
-in that element.
-
-## 3. Reference result
-
-A resolved Surface reference contains:
-
-```text
-path source spelling used by the element
-ref hygienic AuthorValueRef
-type exact nominal TypeRef
-record? defensively copied TypedRecord when the public export is already an authored Record
-```
-
-A component output has no compile-time Record and therefore omits `record`. A Surface that needs to
-inspect a value, such as a Recipe decoder, must require an authored Record and fail during `check`
-when the reference is unresolved, has the wrong Type, or has no value yet.
-
-Only public record exports of an imported SourceUnit are exposed. Private child records are not
-part of the resolver. Every returned Record is a defensive canonical copy so Surface code cannot
-rewrite an already compiled child SourceUnit or a previous declaration.
-
-## 4. Package ownership
-
-For this source:
-
-```svml
-<caption-fine:Style recipe={studio.caption.dialogue}/>
-```
-
-the responsibilities are:
-
-```text
-@narratage/svs
- parses studio.svs
- produces generic @narratage/svs#Recipe
-
-@narratage/markup
- parses the element and the whole-value reference
- resolves only the public imported binding
-
-@narratage/caption Surface
- requires @narratage/svs#Recipe
- validates Caption-owned properties
- emits nominal @narratage/caption#CaptionTrackProgram
- emits the AuthorComponent and locked GraphFragment
-
-Author compiler
- resolves symbolic component dependencies and types
-
-Core
- sees only typed Records, LogicalOutputs, Candidates and Operations
-```
-
-There is no universal property table. Another package may interpret the same generic Recipe through
-another explicitly imported Surface and produce a different nominal Program. The Surface is never
-selected by the Recipe shape; the author selected it by writing the namespaced component.
-
-## 5. Surface output law
-
-A Surface may emit only:
-
-1. authored Records whose exact Type appears in its locked Manifest `outputs` allowlist;
-2. AuthorComponents whose Fragment digest is contributed by that decode result;
-3. immutable GraphFragments;
-4. non-semantic source-map facts.
-
-The Frontend seals authored Record provenance with source and Frontend Closure identity, verifies
-its structural Type, and passes it through the Host admission gate before linking. A Surface cannot
-inject a Record of another undeclared Type.
-
-A typed Program derived from a Recipe should contain any source digest required by its package's
-semantic validation. Its authored provenance already binds the Surface implementation digest and
-the recursively compiled Source Closure. Alias spelling and filesystem location may change source
-identity without changing the typed Program value or resulting Graph meaning.
-
-## 6. Forward-reference law
-
-Two cases are intentionally different:
-
-- Graph dependencies may remain symbolic `component.output` references and are resolved after the
- complete document is collected, so ordinary component forward references remain valid.
-- Compile-time value inspection requires an already compiled authored Record. Imported sources are
- always available; same-document value definitions must appear before a Surface that inspects
- them.
-
-This is not a parser mode or Runtime mode. It follows from whether compilation needs the value now
-or only needs a typed graph edge later.
-
-## 7. What this is not
-
-Surface Binding is not:
-
-- a Core Type registry;
-- an implicit conversion search;
-- a Runtime Producer or paid Operation;
-- a Provider capability;
-- a global SVS decoder;
-- permission to read credentials, files, network or stores.
-
-The current Host executes only explicitly registered trusted Surface implementations. Locked code
-bytes, permission enforcement and isolation are required before arbitrary community Surface code
-can execute.
-
-## 8. Executable proof
-
-The non-video `example.recipe-card` fixture proves the boundary:
-
-1. SVS emits a generic Recipe;
-2. the Card Surface reads only its explicit public reference;
-3. Card validates `fill` and `padding` during author compilation;
-4. Card emits a nominal `CardAppearance` Record;
-5. the runtime Producer consumes `CardAppearance`, not `SVS Recipe`;
-6. an invalid color fails before a BuildPlan or external operation exists;
-7. changing only the source-import alias preserves the Program Record and Graph identities.
-
-Video packages must use this same mechanism rather than adding a video-specific compiler path.
diff --git a/docs/guide/spec/caption-fine.md b/docs/guide/spec/caption-fine.md
deleted file mode 100644
index ded082d1..00000000
--- a/docs/guide/spec/caption-fine.md
+++ /dev/null
@@ -1,239 +0,0 @@
----
-title: Fine Caption Style Family
-description: implemented pre-release Style family with complete browser evidence for the declared surface.
----
-
-# Fine Caption Style Family
-
-Status: implemented pre-release Style family with complete browser evidence for the declared surface.
-
-## Purpose and boundary
-
-`@narratage/caption-fine` is the official field-free, fine-grained Caption Style family. "Fine"
-means that immutable display Atoms are planned into short semantic Cues and then rendered through
-one deterministic visual system. It does **not** mean `important`, semantic emphasis classes or
-random per-word styling.
-
-The package consumes three facts which already exist:
-
-1. a `CaptionProgram`, which assigns one complete Style to every display Word;
-2. a `TimedCaptionProjection`, which gives every Cue and whole Atom a proven time window;
-3. a `CaptionDisplaySequence`, which preserves the author's exact visible text and punctuation.
-
-It emits one ordinary peer `VisualTrack`. It never changes text, calls a model, infers time inside a
-Dual Text Atom or adds Caption metadata to another contract.
-
-| Owner | Responsibility |
-|---|---|
-| `@narratage/caption` | generic Style envelope, Cue/field Plan law, total Style assignment and Atom timing join |
-| `@narratage/caption-fine` | Fine Recipe schema, layout, Paint, local motion and VisualTrack lowering |
-| `@narratage/media` | exact reusable bytes such as `FontArtifactRef`, never Fine typography policy |
-| `@narratage/visual-ir` | closed code-free terminal element/style/keyframe vocabulary |
-| Core | graph demand, validation, derivation and execution only; no Caption meaning |
-
-Changing a Fine shadow, anchor, karaoke mode or Cue box must not modify Caption, Media,
-Composition, Visual IR or Core while their existing public contracts can express the result.
-
-## Complete orthogonal surface
-
-Fine owns one single-font Cue/Atom layout tree. Its dimensions are independent: a Recipe may combine
-gradient glyphs, a current-word box, trail text, an underline and local motion without selecting a
-different renderer. A new package is warranted only when a look changes the layout tree or requires
-planner fields—for example dual-font editorial layout—not when it merely adds Paint or local motion.
-
-### 1. Cue planning
-
-- `cue-min-words`
-- `cue-max-words`
-
-The planner may only cut between whole Atoms. An indivisible author Atom may exceed the preferred
-maximum; it is never split to satisfy a number.
-
-### 2. Placement and layout
-
-- normalized `x`, `y`, `width`
-- `anchor-x: left | center | right`
-- `anchor-y: top | center | bottom`
-- `align: left | center | right`
-- `direction: ltr | rtl`
-- `line-height`, `letter-spacing`, `word-gap`
-- absolute `stack-order`
-
-The Cue naturally wraps only between Atoms. Words inside one Atom never wrap apart. Fine has no
-`max-lines` property: a hard line limit would either discard author text or smuggle browser
-measurement into deterministic lowering. Fine never clips or truncates. Authors control density
-with Cue bounds, width and font size. A future hard layout assertion, if one is genuinely needed,
-must be a separately connected component whose failure is explicit in the graph.
-
-### 3. Typography
-
-- Recipe: `size`
-- `text-transform: none | uppercase | lowercase`
-
-Text transform is terminal presentation only. It never changes `CaptionDisplaySequence`, planner
-input, Word/Atom identity, speech correspondence or timing.
-
-Family, weight and style are deliberately absent from the Recipe. Every Style must import exact
-installed faces from `@narratage/fonts-open`, or declare custom/brand bytes with
-`<media:Font>`. `font=` accepts one exact face with ordered Style-owned
-Fallback children, or one reusable generic `FontStackRef`. The compact open-font path is:
-
-```svml
-<fonts:Stack id="caption-fonts" family="inter" weight="700" style="normal" emoji="color">
- <fonts:Fallback family="noto-sans-sc" weight="700" style="normal"/>
-</fonts:Stack>
-<caption-fine:Style id="primary" recipe={studio.caption.primary} font={caption-fonts}/>
-```
-
-The font edge is the single source of truth for family, weight and style. A fallback preserves its
-own honest face metadata—for example a 700-weight Latin primary may use a 400-weight symbol fallback. Exact
-duplicate faces are rejected. One logical face may contain several content-addressed Unicode-range
-sources, as the installed CJK and Emoji fonts do. Fine expands `FontStackRef` into the same ordered
-exact faces and puts them only
-on its terminal text elements; Caption, Core and unrelated graph values remain unchanged. Omitting
-`font=` is invalid, so no Runtime or machine-font selection policy enters the result.
-
-CJK speech may be authored directly. A display-only emoji still needs explicit speech
-correspondence, for example `<🌐 | globe>`; Script correctly refuses to invent a spoken token for a
-bare symbol. A character with both text and Emoji presentation uses the exact authored Unicode
-sequence (for example `☎️` with VS16); no lowerer rewrites display text to force color. This is
-timing and text truth, not a font limitation.
-
-### 4. Base and active glyph Paint
-
-- solid `fill`, or `gradient-from`, `gradient-to`, `gradient-angle`
-- `opacity`
-- one outline: `stroke-color`, `stroke-width`
-- one shadow: `shadow-color`, `shadow-opacity`, `shadow-x`, `shadow-y`, `shadow-blur`
-- one bounded directional long shadow: `long-shadow-color`, `long-shadow-opacity`,
- `long-shadow-distance`, `long-shadow-angle`
-- one glow: `glow-color`, `glow-opacity`, `glow-blur`
-- one base underline: `underline: off | always`, `underline-color`, `underline-thickness`,
- `underline-offset`
-
-Every active Paint name uses the `active-` prefix. For example `active-gradient-from`,
-`active-long-shadow-distance` and `active-glow-blur` describe the active glyph without changing
-the base glyph.
-
-Base and active Paint have the same shape. A gradient replaces the solid fill for that layer; it
-does not replace the active overlay or the karaoke transition. Drop shadow, long shadow and glow are
-independent authored contributions lowered into one deterministic `text-shadow` declaration. Long
-shadow expansion is implementation-bounded, so one Recipe cannot create an unbounded CSS payload.
-Multiple arbitrary shadow layers, textures, bevel and free-form extrusion remain outside Fine.
-
-### 5. Cue box Paint
-
-- solid `background`
-- `border-color`, `border-width`
-- horizontal/vertical `padding`
-- `radius`
-
-Backdrop sampling is forbidden because it would make this Track inspect another Track's pixels.
-
-### 6. Activation channels, karaoke Paint and timing
-
-- active glyph state: `karaoke: off | current | trail`
-- active glyph transition: `karaoke-transition: step | wipe`
-- active box state: `active-box: off | current | trail`
-- active box continuity: `active-box-continuity: isolated | joined`
-- active box Paint: `active-box-background`, border, padding and radius
-- active box motion: `active-box-enter`, `active-box-exit`, `active-box-transition-frames`
-- active underline state: `active-underline: off | current | trail`, plus color, thickness and offset
-
-Activation state and decoration geometry are deliberately separate. `current` activates only the
-Atom whose measured window contains the frame. `trail` retains every activated Atom through the end
-of the Cue. `step` swaps the whole active glyph layer; `wipe` reveals it across the Atom's measured
-window. RTL reverses the wipe direction. Glyphs, boxes and underlines may choose different state
-policies, so the retained legacy behavior—trail text with a current-only pill—is directly expressible.
-
-An isolated box paints one box per activated Atom. A joined trail paints the one ordered activated
-prefix as continuous inline fragments: atoms on the same rendered line share one background, while
-each wrapped line receives its own end caps. Browser line layout, not upstream metadata, determines
-the fragments. Fine never propagates measured line boxes through the graph.
-
-Karaoke is deliberately Atom-grained. In ordinary text an Atom is normally one visible Word. In
-`<45% | forty five percent>` the authored visible `45%` is one Atom and activates as one unit. In a
-multi-word display Atom, the whole authored Atom activates together. No downstream package invents
-internal Word timestamps that the author and audio evidence never supplied.
-
-### 7. Layered local motion
-
-- Cue enter/exit: `none | fade | pop | scale | spring | bounce | elastic | stamp | tilt |
- zoom-blur | flip-x | flip-y | spin | squash | stretch | slide-left | slide-right | slide-up |
- slide-down | blur-in | wipe-left | wipe-right | wipe-up | wipe-down`
-- Atom entry and exit: the same one-shot vocabulary, with independent durations
-- Atom reveal: `all | on-start | typewriter`
-- active response and active-box enter/exit: the same one-shot vocabulary
-- continuous local loop: `none | shake | wobble | glow-pulse | breathe | float | pulse | flicker`,
- targeted at the Cue or active Atom
-
-The flat Recipe names are `cue-enter`, `cue-exit`, their independent `*-frames`; `atom-enter`,
-`atom-enter-frames`, `atom-exit`, `atom-exit-frames`, `atom-reveal`; `active-response`,
-`active-response-frames`, `active-scale`;
-`slide-distance`; and `loop`, `loop-target`, `loop-period-frames`, `loop-intensity`.
-
-Cue, Atom lifecycle, active response and loop own separate wrappers, so their transforms compose instead
-of overwriting one another. Typewriter reveals authored display graphemes inside a whole Atom; it is
-visual interpolation, not a claim of character timestamps and never changes Atom boundaries.
-Continuous motion is a bounded deterministic frame function with no random input. Arbitrary path
-motion and random variation remain outside Fine; they belong to generic Typography or a separate
-effect package rather than a Caption preset vocabulary.
-
-## One renderer, one layout tree
-
-Static and karaoke output share this structure:
-
-```text
-placement
-└── cue motion
- └── cue box
- ├── joined/isolated decoration underlay
- └── cue loop
- └── atom entry
- └── active response
- ├── base words
- └── active glyph/underline overlay
-```
-
-The decoration underlay repeats only the same transparent glyph geometry required to obtain honest
-browser line fragments. It does not create another text truth. Active overlays reuse the same Atom
-geometry. Every animation lowers to finite frame-addressed `VisualTrack` keyframes before
-HyperFrames sees it.
-
-## Defaults and compatibility
-
-The original fifteen baseline properties remain required, so a minimal Recipe is explicit about
-planning, geometry, typography and its visible box. New dimensions are optional and resolve to a
-complete immutable parameter object:
-
-- top-left anchor, LTR, normal font, zero letter spacing and a one-quarter-em word gap;
-- no stroke, shadow, long shadow, glow, underline or border;
-- solid fill, text transform none;
-- active glyph, active box and active underline off;
-- Cue/Atom/box motion and loops off, all Atoms visible; active response scale defaults to 1.08 when
- that response channel is enabled;
-- when karaoke is enabled without another active Paint, active fill defaults to `#FFD54A` and the
- other active Paint dimensions inherit base Paint.
-
-Unknown properties are rejected. Defaults are package implementation policy and therefore covered
-by its implementation digest; they are not hidden Runtime behavior.
-
-## Completed evidence gates
-
-The complete design was delivered progressively rather than as incompatible versions:
-
-1. **Resolved model and static lowering** — complete parameters, anchors, layout, base/active Paint,
- Cue box and strict validation.
-2. **Timed lowering** — current/trail, step/wipe, Cue fades, Atom reveal and active scale using only
- proven whole-Atom time.
-3. **Reproducibility evidence** — an ordered exact Font Artifact stack, CJK, emoji/symbol fallback,
- multiline wrapping, outline, shadow, glow and all four glyph karaoke modes have real browser/pixel
- witnesses. `max-lines` was deliberately rejected rather than deferred.
-4. **Expressive Paint and motion evidence** — current-only and trail boxes, isolated and joined
- geometry, gradient, underline, long shadow, text transform, layered one-shot motion, typewriter
- reveal and deterministic loops each have a terminal-IR witness; joined geometry additionally has
- a real wrapped browser/pixel witness.
-
-No gate changes Core or common Caption. Exact font selection is an explicit author-graph reference
-between the Media Font Surface and the Fine Style Surface, not metadata propagated through the
-Caption pipeline.
diff --git a/docs/guide/spec/caption-program.md b/docs/guide/spec/caption-program.md
deleted file mode 100644
index 7c9fcda4..00000000
--- a/docs/guide/spec/caption-program.md
+++ /dev/null
@@ -1,191 +0,0 @@
----
-title: Caption Program
-description: pre-release executable contract.
----
-
-# Caption Program
-
-Status: pre-release executable contract.
-
-## 1. Scope
-
-Caption is an ordinary video component family. It turns immutable authored display text into one
-self-contained `VisualTrack`. It has no authority over Script wording, pronunciation, speech
-alignment, Provider selection, Runtime execution or Composition order.
-
-Its two independent graph branches are:
-
-```text
-CaptionDisplaySequence + CaptionProgram ── planner ─────────── CaptionPlan
-speech audio ────────────────────────────── evidence/locator ── CompleteSemanticMap
-
-Display + Correspondence + Plan + Program + Map ── timing ─── TimedCaptionProjection
-TimedCaptionProjection + Display + Program ─────── renderer ── VisualTrack
-```
-
-Planning never receives timestamps. Alignment never receives Cue or Style judgment. The branches
-meet only at the explicit timing component.
-
-## 2. Display truth: Atom and Word
-
-Script emits one `CaptionDisplaySequence` with two ordered views:
-
-- `Atom` is the smallest unit a planner may place wholly into one Cue and the smallest display unit
- that can receive proven speech timing;
-- `Word` is the smallest target for optional Style-owned fields.
-
-Every Word belongs to exactly one Atom. Atoms partition Words exactly once and in order. Text and
-punctuation remain author truth: `45%`, `back-and-forth`, `damn!`, `300,000`, `don't` and `U.S.A.`
-cross this edge unchanged. The official Script reader is English-first and uses authored whitespace
-as its normal display-word boundary; another author package may publish another policy while
-producing the same public contracts.
-
-For ordinary text, one display Word is normally one Atom. Explicit Dual Text is always one whole
-Atom, even when both sides happen to contain matching words:
-
-```svml
-<New York City | new york city>
-```
-
-The display sequence contains one Atom with three Words: `New`, `York`, `City`. It does not claim
-that any one display Word corresponds to any one spoken token. Hidden Dual Text (`< | um>`) produces
-no display Atom.
-
-This is intentional information discipline: Script uses all correspondence the author explicitly
-provided and invents none that the author did not provide.
-
-## 3. Speech correspondence is a separate edge
-
-`CaptionCorrespondence` maps every display Atom to one or more authored speech-token identities.
-It is separate from `CaptionDisplaySequence`; display consumers do not carry speech lineage they do
-not need.
-
-For ordinary text, each Atom maps to the speech tokens structurally owned by that display surface.
-For Dual Text, the whole display Atom maps to the whole right-side speech range. There is no public
-display-Word-to-speech-Word mapping and no guessed internal timing.
-
-Script also emits `CaptionDisplayWordSubset` values for explicit Selections. Role selectors are
-author-surface sugar over the same display sequence. A subset is ordered and may select zero or
-more whole Atoms; selecting only part of a multiword Atom is an authoring error. These values are
-not temporal `NarrativeSelection` values and do not add another `when` language.
-
-## 4. Complete Style intent
-
-A `CaptionStyle` is a complete author declaration with:
-
-- universal planning requirements: preferred Cue word bounds and zero or more typed per-Word field
- declarations;
-- one rendering-family identity and that family's complete Recipe-resolved parameters.
-
-The common Caption package owns the Cue and field data shapes, not fonts, boxes, colors, animation
-or the meaning of a field id. A Style family may declare no fields. Boolean, enum and bounded-number
-fields are available to other families; a Word may carry zero, one or several independent fields.
-
-SVS stores named Recipe data. A Style package interprets that Recipe and emits a complete Style.
-Applying another Style replaces the whole Style; arbitrary planning and rendering pieces are never
-implicitly merged.
-
-## 5. Total assignment and ordered replacement
-
-Every `CaptionProgram` has one explicit default Style covering the complete display universe.
-Authors do not need a fake `@whole` Selection or a complement.
-
-```xml
-<caption:Program id="captions" display={story.caption} default={plain}>
- <caption:Use role="ALICE" style={alice}/>
- <caption:Use words={story.caption.selection.special} style={impact}/>
- <caption:Mute words={story.caption.selection.private}/>
-</caption:Program>
-```
-
-Uses apply in source order and the last matching whole-Style replacement wins. The Program stores
-the display-sequence identity and ordered Word-id runs, not copied text. A run never crosses a
-Segment or Turn boundary. All runs are disjoint and partition the display Words exactly once, and
-no run may split an Atom.
-
-`Mute` consumes the same Caption-specific whole-Atom word subset as `Use`. Multiple `Mute`
-children form one ordered union, so one Program can hide any number of disconnected authored
-ranges. `Mute` is post-planning visibility intent: muted Atoms remain in the immutable display
-universe and `CaptionPlan`, never enter the planner prompt as an instruction, and never cause Cue
-replanning. The common timing join removes muted Atoms from the already-planned timed Cue while
-preserving that Cue's identity and original time span; a Cue with no visible Atom is absent from the
-timed visual projection. Style-family renderers therefore share this behavior without each
-inventing their own mask semantics. The common package also exports the same idempotent operation
-for defensive direct renderer calls.
-
-This is deliberately not a generic temporal mask. Script has already projected the author's
-Selection into the exact display-word subset, so Caption need not convert words to time and then
-guess them back from overlapping windows. A `Mute` selecting only part of a multiword Dual Text
-Atom is rejected by the same indivisibility rule as a Style application.
-
-`display={story.caption}` is a real graph edge. Planner, timing and rendering declare their own
-edges to the same value only when they need it.
-
-## 6. Planner contract
-
-A planner receives resolved runs. For each run it may only:
-
-1. group consecutive whole Atoms into ordered Cues;
-2. assign declared fields to Words inside each Cue.
-
-Cues must partition every run exactly once. Preferred minimum/maximum Cue word counts guide the
-planner but cannot invalidate author text; a single multiword Atom may exceed the maximum and must
-remain whole. The response contains no replacement text, pronunciation, timestamp or Style.
-
-The Gemini planner sends the readable form only once:
-
-```json
-{
- "atoms": [
- ["new"],
- ["york"],
- ["city", "is", "beautiful"]
- ]
-}
-```
-
-The outer array is the Cue-cut universe. Each inner array is one indivisible Atom. Strings inside
-it are immutable field targets. Stable ids remain inside the package. A response consumes Atoms by
-`atom_count`; a field uses one-based `atom_number` and `word_number` within its Cue.
-
-## 7. Timing projection
-
-`CompleteSemanticMap` remains the only global speech timing truth. The timing component resolves
-each Atom's explicit speech-token range through `CaptionCorrespondence` and gives that whole Atom
-one measured window. A Cue begins at its first Atom and ends at its last Atom.
-
-`TimedCaptionProjection` therefore contains timed Cues and timed Atoms only. It contains no
-per-display-Word timestamps, no `estimated` quality label and no proportional character split. For
-a multiword Dual Text Atom, downstream renderers may choose how to lay out or animate its internal
-Words, but they cannot claim invented word timing. Nothing is written back into SemanticMap.
-
-## 8. Rendering and composition
-
-Each timed Cue selects the complete rendering payload of its resolved Style. One Style-family
-renderer returns one ordinary `VisualTrack`. In the first contract, all Styles in one Program use
-the same rendering family.
-
-`@narratage/caption-fine` is the first official family. It declares no fields and owns one
-orthogonal single-font layout/Paint/motion system, including independent whole-Atom glyph,
-underline and Pill activation. Another package may define a different layout tree or field-dependent
-Paint without modifying Script, Caption, Composition or Core.
-
-Film does not recognize Caption specially. It folds the resulting peer Track beside Speech,
-B-roll, Text and third-party Tracks.
-
-## 9. Explicit non-goals
-
-Caption v1 does not define:
-
-- Script correction or alternative wording;
-- inferred Dual Text word correspondence;
-- speaker diarization or speaker identity;
-- implicit model or Provider routing;
-- automatic cache reuse;
-- estimated display-word timing;
-- planner-visible or Core-owned Caption mute semantics;
-- a global Caption layer in Composition;
-- field-specific semantics in Core.
-
-Targets and Candidates remain external run intent. They do not change Caption data or
-planner semantics.
diff --git a/docs/guide/spec/core-kernel.md b/docs/guide/spec/core-kernel.md
deleted file mode 100644
index 0061681a..00000000
--- a/docs/guide/spec/core-kernel.md
+++ /dev/null
@@ -1,87 +0,0 @@
----
-title: Narratage Core Kernel
-description: current executable contract. Persisted Graph, BuildRequest, Plan and BuildState wire.
----
-
-# Narratage Core Kernel
-
-Status: current executable contract. Persisted Graph, BuildRequest, Plan and BuildState wire
-formats are `@1`.
-
-Core is a domain-free graph-demand compiler and verified Build state machine. It does not parse
-source, load packages, execute implementations, choose an Endpoint, access credentials, persist
-artifacts or know video vocabulary.
-
-## Inputs
-
-Core accepts immutable, already linked data:
-
-- a resolved module closure and typed authored Records;
-- a complete `svml.graph@1` containing Logical Outputs, independent Candidates and atomic
- Operations from the Author and Run Graphs;
-- a `svml.build-request@1` containing Targets.
-
-Graph construction, source parsing, package resolution, Fragment expansion and Run Graph authoring
-happen outside Core. Core verifies the resulting data before deriving a plan.
-
-## Graph values
-
-1. A `LogicalOutput` is a stable author-visible promise: identity, nominal Type and Primary
- Candidate.
-2. A `Candidate` is an independent typed supply. Its root is one immutable Provided Value or one
- atomic Operation result. It does not belong to an output.
-3. The Run compiler applies each explicit `Satisfaction` by making its Candidate that output's
- Primary Candidate in the immutable realized graph. Core receives no second selection source.
-4. A `Target` names an output required by this Build.
-5. An `Operation` has stable instance identity, declared typed inputs and exactly one atomic result.
- Multi-result components use one Product followed by deterministic Projection Operations.
-
-## Compilation laws
-
-1. Core derives Demand backwards from all Targets through selected Candidates and Operation inputs.
-2. Selection is already frozen into the graph; reachability alone determines work.
-3. One Operation instance referenced by several edges is included once. Different instances are
- never content-deduplicated, even when implementation, parameters or input Records match.
-4. A Provided-Value Candidate has no input edge, so traversal stops naturally. Core has no Pin,
- preview, cache or historical-result branch.
-5. A Candidate must produce the exact nominal Type required by the Logical Output using it.
-6. Cycles, missing references, undeclared ports, duplicate identities and incompatible Types fail
- before execution.
-7. The resulting `svml.plan@1` is finite, deterministic and content-bound to the verified Graph and
- BuildRequest.
-
-## Execution laws
-
-1. Core emits Commands from the frozen BuildPlan. Driver and Runtime cannot change Candidate
- selection, dependency topology or Target meaning.
-2. Records, Needs, Commands, Events, Receipts and Derivations are immutable content-bound facts.
-3. A Derivation binds implementation identity, input and output digests, Need request digests and
- the accepted Event digest.
-4. An external Receipt binds the exact Need, request digest, fulfillment implementation, output
- digest and accepted Event.
-5. Serialized outstanding Commands are not trusted. Resume verifies accepted state, discards old
- commands and deterministically regenerates the next commands.
-6. A Build completes only when every Target Record exists.
-
-## Type laws
-
-1. Types are nominally owned by modules, not registered in a central Core union.
-2. Core checks the locked structural Schema for every admitted Record.
-3. A Type owner may lock a semantic validator digest. When present, every Record of that Type needs
- a receipt bound to the exact Type, value digest and validator implementation.
-4. The Graph owns dependencies, Derivation owns deterministic lineage and Receipt owns external
- fulfillment provenance.
-
-## Boundary
-
-`@narratage/protocol` owns wire data and identity. `@narratage/core` owns verification, Demand compilation and
-state transitions. Compiler packages own source and graph construction. Runtime owns scheduling,
-persistence and placement. Endpoint packages fulfill exact Capabilities. Domain packages own shared
-vocabulary and validators.
-
-Neither a file suffix, package name nor TypeScript class grants Kernel privilege. A domain package
-released after Core can participate through manifests, nominal Types and ordinary Operations without
-a Core release.
-
-The Kernel has no privileged terminal output, Film root, video pipeline direction or single-Target
-restriction.
diff --git a/docs/guide/spec/deck-track.md b/docs/guide/spec/deck-track.md
deleted file mode 100644
index 311cdff5..00000000
--- a/docs/guide/spec/deck-track.md
+++ /dev/null
@@ -1,291 +0,0 @@
----
-title: SVML Depth-Stack Deck Track
-description: implemented executable authority for the official depth-stack Deck migration. It records.
----
-
-# SVML Depth-Stack Deck Track
-
-Status: implemented executable authority for the official depth-stack Deck migration. It records
-the behavior used by the old Deck implementation, but it is not a frozen public ABI and does not
-define a universal Deck protocol.
-
-## 1. Conclusion
-
-A Deck is not a Media Item mode and is not the third primitive of Media Track. It is a higher-order
-visual component whose members happen to use media material.
-
-The first official implementation is `@narratage/deck-track`. Its honest author
-meaning is a **depth-stack Deck**:
-
-```text
-ordered Cards
-semantic activation points
-current Card
-visible previous/future neighborhood
-relative-depth layout
-whole-collection state transition
-ordinary VisualTrack output
-```
-
-It is a peer of Media Track, Text, Caption and Ranking:
-
-```text
-Media Item / Sequence ─┐
-Depth-Stack Deck ──────┤
-Text ──────────────────┼─> ordinary VisualTrack values ─> Composition
-Caption ───────────────┤
-Ranking ───────────────┘
-```
-
-Core, Film, Composition and HyperFrames do not know that a contribution came from a Deck.
-
-## 2. Why Deck is outside Media Track
-
-A Media Item answers how one material appears. A Media Sequence answers how members replace one
-another on one shared surface. A depth-stack Deck answers a different question: how an entire
-ordered collection changes pose when its current member changes.
-
-At one frame a Deck may simultaneously show:
-
-- the current Card;
-- several retained historical Cards;
-- several future preview Cards;
-- different offsets, scales, rotations, opacity and tone by relative depth;
-- a transition in which every retained Card moves from its old pose to its new pose.
-
-That is collection state, not source sampling, fitting or an outgoing/incoming pair handoff. Keeping
-it inside Media Track would make one historical presentation model look like a universal media
-primitive.
-
-## 3. No universal Deck contract yet
-
-“Deck” is currently a design family, not a shared public Type. Other useful models may have
-fundamentally different topology:
-
-- horizontal or circular Carousel;
-- fan layout;
-- Cover Flow;
-- grid with a distinguished current member;
-- freeform collage;
-- time-axis history;
-- a Ranking board whose board and entries evolve together.
-
-Those packages may define different author Surfaces, validation and lowering. They need only emit
-the terminal `VisualTrack` contract. We should extract a shared Deck library only after multiple
-real implementations prove a smaller common law. Merely sharing the noun “Deck” is not evidence for
-a central protocol.
-
-Consequently the first package should expose an honest component name such as `DepthStack`, rather
-than claiming that its fields describe every possible Deck.
-
-## 4. Package boundary
-
-The intended relationship is:
-
-```text
-@narratage/temporal point location and exact trigger schedule
-@narratage/spatial Canvas, Placement Frame and fitted geometry
-@narratage/media normalized still/timed/Surface material
-@narratage/media-track focused reusable media-layer lowering implementation
-@narratage/deck-track DepthStack author Program, state and reflow lowering
-@narratage/composition terminal VisualTrack waist
-```
-
-`@narratage/deck-track` may reuse focused implementation functions from Media Track or a later
-implementation-only media-lowering library. That code reuse does not make Deck a Media Track
-component, and Core exposes no universal media or deck program.
-
-The package does not:
-
-- generate Card images or videos;
-- select Providers or credentials;
-- infer Cards from URLs, directories or media metadata;
-- inspect another Track or the accumulated lower composite;
-- reserve a global stacking band;
-- add Deck meaning to Core, Film, Composition or HyperFrames.
-
-Every Card source, label, trigger and sound dependency is an explicit graph reference.
-
-## 5. Illustrative author Surface
-
-The syntax is not frozen, but package ownership should be visible:
-
-```xml
-<deck:DepthStack
- id="proof-stack"
- map={timing.map}
- space={speech.space}
- canvas={vertical}
- frame={layout.lower-proof-stack}
- until={story.selection.proof.end}
- appearance={studio.deck.proof}
->
- <deck:Card id="proof-1" source={proof1.image} at={story.moment.proof1}/>
- <deck:Card id="proof-2" source={proof2.image} at={story.moment.proof2}/>
- <deck:Card id="proof-3" source={proof3.video} at={story.moment.proof3}/>
-</deck:DepthStack>
-```
-
-The component itself produces an ordinary `VisualTrack`. It is not nested in `<media:Track>` and it
-is not selected through `mode="deck"`.
-
-SVML owns topology: ordered Card membership, source edges, trigger references, terminal point,
-Frame and optional label inputs. SVS owns reusable parameters: visible neighborhood, depth poses,
-frame Paint and reflow motion. SVS cannot create Cards, hide source edges or choose a Provider.
-
-## 6. Exact state model
-
-Let authored Card activation frames be:
-
-```text
-p1 < p2 < ... < pN < T
-```
-
-where `T` is the explicit terminal point. At frame `f`:
-
-```text
-current(f) = greatest i such that pi <= f
-```
-
-Before `p1`, the component is outside its active state unless an explicit initial-state policy says
-otherwise. At and after `T`, it is outside the active state except for its explicitly authored exit
-motion.
-
-The package validates authored occurrence order. It never sorts equal, reversed or missing trigger
-points into a seemingly valid presentation.
-
-Visibility is an explicit package parameter:
-
-```ts
-type DepthStackVisibility = {
- readonly previous: number;
- readonly next: number;
- readonly wrap: boolean;
-};
-```
-
-Every visible Card receives an integer relative depth around `current(f)`. A pure layout function
-maps that depth to its pose:
-
-```ts
-type DepthStackPose = {
- readonly x: number;
- readonly y: number;
- readonly scale: number;
- readonly rotationDeg: number;
- readonly opacity: number;
- readonly stacking: number;
- readonly tone?: DeckCardTone;
-};
-```
-
-The public author model should expose named pose/visibility Recipes, not require authors to repeat
-the complete table inline.
-
-## 7. Reflow is a collection transition
-
-When Card `i` becomes current, every Card visible in either the old or new state obtains an old pose
-and a new pose. The reflow transition interpolates between those poses with one exact duration and
-one easing law.
-
-This differs from a Media Sequence handoff:
-
-```text
-Sequence transition coordinates outgoing member + incoming member
-Deck reflow coordinates the union of old-visible + new-visible Cards
-```
-
-They must not share one `transitionStyle` union. A Card may additionally have owned current-card
-emphasis, but it cannot mutate pixels belonging to another Track.
-
-The renderer evaluates state and pose as pure functions of the absolute frame. It does not retain
-mutable “previous frame” state. Therefore arbitrary-frame rendering, chunked Lambda rendering and
-out-of-order frame evaluation remain deterministic.
-
-## 8. Card material and playback
-
-Each Card may reuse the same explicit layer, Content Fit, clipping and frame-Paint laws as a Media
-Item. That is implementation reuse, not author-model inheritance.
-
-Timed Card sources must define how they sample while not current:
-
-```text
-future preview hold source head, or sample an explicitly selected visible clock
-current Card sample according to its authored active-phase occupancy
-past trail hold source tail, continue, or hide according to explicit policy
-```
-
-Still material needs no playback policy. A video must never silently finish while it is only a
-future preview.
-
-Card labels are optional explicit Text inputs. The Deck package may reuse typography lowering, but
-it cannot derive Text truth from filenames, URLs or hidden image metadata.
-
-The initial DepthStack is visual-only. If a later Deck component exposes source audio, it must do so
-as a separate explicit `AudioTrack` projection with an authored current-member/handoff law; audio
-must never become audible merely because a Card's container contains a stream.
-
-## 9. Stacking
-
-Deck Track has no track-level z-index and creates no stacking context. The board/background and
-individual Cards may lower to separate absolute-stack Presents when peer Tracks need to interleave
-between them.
-
-Relative depth determines only ordering among owned Cards. Temporary ordering during reflow exists
-only inside that explicit relationship. It does not allocate a global z band.
-
-## 10. Migration from the legacy implementation
-
-Retain from the old Deck implementation:
-
-- explicit authored Card order;
-- semantic trigger-driven current state;
-- configurable previous/future visibility;
-- finite or explicit wrapping behavior;
-- depth offset, scale, alternating rotation, opacity and tone;
-- whole-group entry, sustain and exit motion;
-- full-group interpolated reflow;
-- frame Paint, padding, radius, border and shadow;
-- deterministic frame-local evaluation.
-
-Retire:
-
-- treating Deck as a B-roll or Media mode;
-- implicit `linkNext` topology;
-- placement semantics hidden behind “usually below the face”;
-- automatic labels from source metadata;
-- lower-composite transitions;
-- lane-global sound defaults;
-- any registry entry in Core, Film, Composition or HyperFrames.
-
-## 11. Implementation and acceptance order
-
-1. implement the shared Temporal and Spatial slices;
-2. expose `DepthStack` and explicit ordered Cards;
-3. resolve exact trigger state and finite visibility;
-4. reuse focused media-layer lowering for Card contents;
-5. add pure relative-depth pose resolution;
-6. add collection reflow and whole-group lifecycle motion;
-7. prove still and timed inactive-playback behavior;
-8. migrate one real old-system Deck and freeze only after the acceptance matrix passes.
-
-Steps 1–8 execute in `@narratage/deck-track`. Package tests cover every state, validation,
-playback, label and extension case below. The real HyperFrames witness renders a legacy-style
-three-Card depth stack twice—one worker and three partitioned workers—and compares every decoded
-frame byte-for-byte while also proving each authored current state paints. This completes the Deck
-package acceptance matrix, and the independent generic E1–E8 terminal VisualTrack/Visual IR gates
-now pass as well.
-
-Required evidence includes:
-
-- previous-only, next-only and mixed neighborhoods;
-- zero, one and many visible neighbors;
-- finite boundaries and explicit wrapping;
-- missing, equal and reversed trigger rejection;
-- depth pose, alternating rotation, opacity and tone;
-- cut and interpolated whole-collection reflow;
-- still and timed Card sampling laws;
-- optional exact-font labels;
-- whole-group entry/sustain/exit without overwriting Card poses;
-- byte-identical frames under sequential, partitioned and out-of-order rendering;
-- installation of another Deck package without changing Core or the first DepthStack package.
diff --git a/docs/guide/spec/media-track.md b/docs/guide/spec/media-track.md
deleted file mode 100644
index 50151a14..00000000
--- a/docs/guide/spec/media-track.md
+++ /dev/null
@@ -1,725 +0,0 @@
----
-title: SVML Media Track Authoring
-description: executable pre-release authority for the unified official Media Track. Item/Sequence.
----
-
-# SVML Media Track Authoring
-
-Status: executable pre-release authority for the unified official Media Track. Item/Sequence
-Surfaces, timed material, ordered layers, motion, handoffs, separate audio projection and the
-restricted Speech projection are implemented. The former `@narratage/broll` vertical slice has
-been retired rather than preserved as compatibility. The contract is not a frozen public ABI.
-
-## 1. Conclusion
-
-Full-canvas cutaways, lower evidence cards, corner images, transparent stickers, animated images,
-short videos and ordinary replacement sequences are not different terminal Track families. They are
-different compositions of the same media concerns:
-
-```text
-source material
-target time / trigger schedule
-source-time occupancy
-Placement Frame
-ordered local layers and Content Fits
-clip and frame Paint
-entry / sustain / exit motion
-per-layer sampling motion
-absolute stacking
-optional owned pairwise replacement relationship
-optional explicit audio projection
-```
-
-They belong in one video-domain package, `@narratage/media-track`, and all lower to
-ordinary peer `VisualTrack` and, only when explicitly authored, `AudioTrack` values.
-
-One package does not mean one conditional super-component. The author Surface exposes two
-components with different meanings:
-
-```text
-Media Item one independent presentation with one projected window
-Media Sequence ordered members replace one another on one shared surface
-```
-
-`Item` and `Sequence` share focused value families and lowering code. They are separate components
-because they have different topology, validation and ports. There is no `mode="broll | sequence |
-overlay"` field and no step where choosing a mode causes a different hidden component to appear.
-
-A depth-stack Deck is deliberately outside this package. It is a higher-order collection-state
-component, not a media primitive. The official migration model is specified independently in
-[`deck-track.md`](./deck-track.md); future Carousel, Fan, Grid or other Deck packages need not adopt
-that model.
-
-“B-roll” remains a useful editorial description or Recipe name. It is not a public data Type.
-
-## 2. Boundary
-
-The unified package is an author package and deterministic compiler:
-
-```text
-@narratage/media intrinsic media facts and normalized sources
-@narratage/media-pipeline inspection, stream selection and normalization vocabulary
-@narratage/temporal points, occurrences, windows and trigger schedules
-@narratage/spatial Canvas, Placement Frame, Content Fit and fitted geometry
-@narratage/media-track Item / Sequence author Programs and lowering
-@narratage/deck-track independent depth-stack Deck authoring and lowering
-@narratage/composition terminal VisualTrack / AudioTrack waist
-```
-
-It does not:
-
-- generate images or videos;
-- choose a model or Provider;
-- inspect credentials, URLs or local executables;
-- let a Runtime or Provider invent stream-selection policy;
-- own the Program clock or Canvas;
-- read a sibling Track or the accumulated lower composite;
-- add a B-roll, Sequence, Deck or Speech branch to Core, Film, Composition or HyperFrames.
-
-Core sees fixed graph Operations, Records and edges. Runtime executes any explicit media
-inspection/normalization or Surface-materialization Needs. Providers do not reinterpret placement,
-motion, transitions or audio policy.
-
-## 3. Source truth
-
-Every visual sample enters through an explicit graph edge. The author Surface distinguishes four
-forms rather than guessing from one Blob type:
-
-- still image material with a content-addressed Artifact and exact display-oriented extent;
-- generated/raw video material, lowered through the generic Media Pipeline using the Track's
- ProgramSpace rate and an explicit visual-only or A/V policy;
-- normalized timed visual material with exact frame rate, frame count and extent;
-- typed `CompositableSurface` material with exact still/frame timing and alpha mode.
-
-The intrinsic owner remains `@narratage/media`. The Media Track package may use a package-local
-resolved union internally, but it must not create another universal media Type or copy provider,
-source path, Narrative digest or upstream lineage into every item.
-
-Filename inference is invalid. In particular:
-
-- an animated GIF or animated WebP cannot remain a browser-timed `<img>` because frame `n` must be
- independently renderable during parallel rendering;
-- it must be normalized into exact frame-timed visual material or a typed animated Surface;
-- a still image remains durationless even when its file format also supports animation;
-- an MP4 containing audio does not make that audio audible;
-- alpha-bearing visual material must retain an explicit alpha contract.
-
-Several layers may reference the same Artifact. The graph and Artifact collector deduplicate bytes;
-the renderer still evaluates every authored sample independently.
-
-## 4. The orthogonal item model
-
-Conceptually, one independent Item owns:
-
-```ts
-type MediaItemProgram = {
- readonly id: string;
- readonly temporal: TemporalBinding;
- readonly frame: SpatialFrameInput;
- readonly framePresentation: MediaFramePresentation;
- readonly layers: readonly MediaLayerProgram[];
- readonly motion: MediaLifecycleMotion;
- readonly stacking: AbsoluteStackingIntent;
- readonly sourceAudio?: MediaSourceAudioProjection;
-};
-```
-
-The axes are:
-
-| Axis | Owns | Does not own |
-|---|---|---|
-| source | exact still/timed/Surface material | Provider or source discovery |
-| temporal | occurrence expansion and item window | source playback |
-| frame | Placement Frame | visible Paint or content aspect |
-| layers | ordered Paint/media samples | Track-relative z context |
-| content fit | one Content Frame per sample | zoom/pan or playback |
-| occupancy | source trim and source-time mapping | target window |
-| frame presentation | clip, radius, border, shadow, padding | geometry construction |
-| lifecycle motion | entry, sustain and exit of owned pixels | member-to-member handoff |
-| sampling motion | pan/zoom/rotation inside one media layer | Placement Frame motion |
-| stacking | absolute Present position | Track family priority |
-| source audio | explicit projection of one selected layer's audio | automatic container audio |
-
-These are package Program facts. They do not become fields on Core or the terminal Track contract.
-
-## 5. Illustrative author Surface
-
-The final Surface must favor named Frames and SVS Recipes over large inline bags. One possible
-shape is:
-
-```xml
-<media:Track id="editorial-media" map={timing.map} space={speech.space} canvas={vertical}>
- <media:Item
- id="full-cutaway"
- video={demo.video}
- during={story.selection.demo}
- frame={vertical}
- appearance={studio.media.full-cutaway}
- />
-
- <media:Item
- id="corner-loop"
- surface={reaction.surface}
- at={story.moment.reaction}
- for="2s"
- frame={layout.top-right-sticker}
- appearance={studio.media.corner-loop}
- />
-
- <media:Sequence
- id="steps"
- frame={layout.lower-card}
- appearance={studio.media.evidence-sequence}
- until={story.selection.explanation.end}
- >
- <media:Member id="step-1" image={step1.image} extent={step1.extent} at={story.moment.step1}/>
- <media:Member id="step-2" image={step2.image} extent={step2.extent} at={story.moment.step2}/>
- <media:Member id="step-3" video={step3.video} at={story.moment.step3}/>
- <media:Handoff from="step-1" transition={studio.media.transitions.push-left}/>
- <media:Handoff from="step-2" transition={studio.media.transitions.crossfade}/>
- </media:Sequence>
-
-</media:Track>
-```
-
-This syntax is illustrative. The normative points are:
-
-- declarations and reusable appearance stay outside usage sites;
-- every source, Frame, Selection and Moment is an explicit reference;
-- `image`, `video`, `media` and `surface` name mutually exclusive source forms; a generic Blob is
- never guessed to be a still or moving source;
-- `Item` and `Sequence` are visibly different author meanings;
-- shorthand such as `during`, `at`, `for` and `until` lowers to the one shared Temporal algebra;
-- no runtime chooses whether something is a cutaway, card or sticker.
-
-### SVML structure and SVS parameter bundles
-
-SVML owns topology: sources, Frames, timing references, Item/Sequence membership, Handoff edges,
-labels and sound inputs. SVS owns named reusable parameter bundles: Content Fits, frame Paint,
-clipping, motion parameters and transition appearance.
-
-An SVS Recipe cannot hide a source edge, create a member, select a Provider or cause two independent
-Items to become a Sequence. Applying one transition Recipe to a whole Sequence materializes that
-choice on every authored Handoff in the resolved Program. Thus Recipes stay concise without
-becoming a second graph language.
-
-### Small framed inset with a smooth rise
-
-A small box is not another component. It is one ordinary Item whose explicit Placement Frame is
-smaller than Canvas. Structure remains readable at the use site:
-
-```svml
-<space:AnchoredFrame id="product-inset" within={safe}
- x="50%" y="78%" width="82%" height="28%" anchor="center"/>
-
-<media:Track id="proof" map={timing.map} space={speech.space} canvas={vertical}>
- <media:Item
- id="product-proof"
- video={product.video}
- during={story.selection.proof}
- frame={product-inset}
- appearance={studio.media.fuzzy-card}
- motion={studio.motion.smooth-rise}
- />
-</media:Track>
-```
-
-The referenced Recipes conceptually resolve independent parameters such as:
-
-```text
-media.fuzzy-card
- content fit contain or cover
- frame fill explicit color/gradient/transparent Paint
- content padding exact pixels
- frame clip rounded rectangle or owned shape
- edge treatment border/shadow, or an explicit rough-edge layer
-
-motion.smooth-rise
- enter operator translate
- from below the Canvas or an exact positive y offset
- to resolved Placement Frame pose
- duration exact frames
- easing ease-out, or a package-compiled custom frame curve
-```
-
-“Below” must be unambiguous. `below-canvas` derives an off-canvas starting pose from Canvas and the
-resolved Item extent; `offset-y: 120px` starts below the final Frame without necessarily leaving the
-picture. Both lower to frame-exact local transforms and neither changes the final Spatial Frame.
-
-A normal soft or glowing edge lowers to an owned border plus box shadow/drop shadow. A genuinely
-irregular furry, torn-paper or noisy edge is not a magic `border-style`: it is an explicit local
-layer, such as a transparent frame Artifact or a deterministic `CompositableSurface`. Because that
-layer affects graph topology and Artifact collection, its source must be connected in SVML (for
-example as `<media:Layer surface={fuzzyFrame.surface}/>`); SVS may style it but cannot hide the source
-edge. The outer lifecycle wrapper moves the frame Paint, edge layer and content together.
-
-Built-in easing names may lower directly to Visual IR. A richer Bézier or spring Recipe remains a
-Media-package concern and may be deterministically sampled into frame-addressed keyframes; it does
-not require Core, Composition or Runtime to learn a new motion type.
-
-## 6. Independent Item timing
-
-An Item uses the common pipeline from [`track-authoring.md`](./track-authoring.md):
-
-1. locate Selection, Moment or Program points;
-2. expand `one` or `each` occurrences;
-3. project one target window per occurrence;
-4. intersect and validate against ProgramSpace;
-5. evaluate every layer's source trim and occupancy inside that item window;
-6. evaluate lifecycle and sampling motion;
-7. emit self-contained Presents.
-
-Independent Items remain independent. Their windows may overlap, and they compose by their explicit
-absolute stacking keys. Program order does not make them exclusive, clip a lower-priority item,
-auto-stitch gaps or infer a Sequence.
-
-A full-screen B-roll is simply an Item whose Placement Frame equals Canvas and whose foreground
-layer commonly uses `cover`. A small image, GIF or video is the same Item against a smaller Frame.
-“Do not cover the face” is not a Media mode: the author consumes a suitable named Frame, or an
-upstream subject-aware layout component explicitly produces one.
-
-## 7. Ordered local layers
-
-One Item owns an ordered list of local layers inside one Placement Frame:
-
-```ts
-type MediaLayerProgram =
- | MediaFramePaintLayer
- | MediaSampleLayer;
-
-type MediaSampleLayer = {
- readonly id: string;
- readonly source: MediaSourceInput;
- readonly fit: ContentFit;
- readonly trim?: VisualSourceTrim;
- readonly occupancy: VisualOccupancy;
- readonly appearance: MediaSampleAppearance;
- readonly samplingMotion?: MediaSamplingMotion;
-};
-```
-
-The complete backing cases from [`spatial-layout.md`](./spatial-layout.md) are ordinary layer lists:
-
-- transparent: no backing layer;
-- solid or gradient: one frame-Paint layer;
-- self blur: the same source enters one cover/blur layer and one foreground layer;
-- alternate backing: a different explicit source supplies the backing layer;
-- decorated card: Paint, border/shadow/padding and one or more media layers.
-
-Every media layer derives its own Content Frame from its own intrinsic extent and `ContentFit`.
-There is no special foreground/background pair and no one shared inner box.
-
-Frame-level clipping is explicit: none, Placement Frame or an owned shape. Radius, border, shadow,
-padding and frame Paint are presentation. They cannot mutate Spatial geometry or sample another
-Track.
-
-## 8. Source-time occupancy
-
-Static material has no intrinsic playback duration and paints for the complete interval in which
-its layer is active. A timed video, animated image or animated Surface uses explicit trim followed
-by the common visual occupancy law:
-
-```text
-once/start play from the source head, then disappear or truncate
-once/end align the source tail to the window end
-hold/start play from the head, then hold the last visual frame
-hold/end hold the first visual frame, then play to the end
-loop/start loop from source-head phase
-loop/end choose phase so the source tail meets the window end
-stretch map the complete effective source interval onto the window
-```
-
-There is no `auto`, `native`, `finish`, `freeze` or `phased` author enum. Images do not receive a
-fake playback mode. Source trim executes before occupancy. Spatial contain/cover and temporal
-stretch remain unrelated.
-
-The Item target window remains valid even when one `once` layer ends early. Other Paint or media
-layers may continue. An author who wants the complete item to remain visible chooses `hold`, supplies
-a persistent layer or authors a shorter target window; the compiler never shrinks the Item by
-guessing which layer is primary.
-
-## 9. Media motion has four separate channels
-
-Old B-roll behavior becomes four explicitly separated package-owned channels:
-
-```ts
-type MediaLifecycleMotion = {
- readonly enter?: MediaEdgeMotion;
- readonly sustain?: readonly MediaSustainMotion[];
- readonly exit?: MediaEdgeMotion;
-};
-
-type MediaSamplingMotion = {
- readonly keyframes: readonly MediaSamplingKeyframe[];
-};
-```
-
-### Entry motion
-
-Operates on all pixels owned by one Item or Group at the beginning of its visible envelope.
-Reasonable typed Recipes include fade, directional slide, scale, pop, bounce, local blur reveal,
-wipe, flip and spin.
-
-### Sustain motion
-
-Operates during the visible envelope. Float, breathe, pulse, wobble, shake and drift are reasonable
-operators. The authored ordered list is evaluated deterministically; multiple transforms cannot
-silently overwrite one another.
-
-### Exit motion
-
-Operates on all owned pixels at the end of the envelope. It uses its own operator and duration; it
-is not automatically the reverse of entry.
-
-### Sampling motion
-
-Operates on one media layer's already fitted Content Frame. Normalized keyframes may animate zoom,
-source/content-point displacement and rotation. Ken Burns is a sampling-motion Recipe, not a Base
-effect and not an occupancy mode.
-
-Entry and exit keep their authored durations even when they overlap. The package evaluates both at
-each frame and deterministically composes their independent opacity, transform, filter and clip
-channels. It never proportionally shortens either side and never runs the Item twice. The visible
-Item window remains the final clip.
-
-The transform stack is fixed so channels cannot overwrite one another:
-
-```text
-absolute Present stacking
-└─ fixed Placement Frame
- └─ group/item entry + exit wrapper
- └─ ordered sustain wrappers
- └─ member handoff wrapper (Sequence only)
- └─ frame Paint and clip
- └─ base fitted Content Frame
- └─ layer sampling motion
- └─ media pixels
-```
-
-The lowerer may algebraically compose matrices or emit nested Visual IR elements, but the order and
-pixels must be identical. These are Media package contracts, not a generic public `effects` bag.
-
-## 10. Sequence: explicit member replacement
-
-A Sequence owns two or more nested Members that share one Placement Frame and one outer lifecycle.
-It is created explicitly; nearby Items, overlapping windows or a transition Recipe never imply
-membership.
-
-Each Member has:
-
-- a stable author identity;
-- explicit media layers and per-layer fit/occupancy/sampling motion;
-- exactly one activation point derived from a Moment or one Selection boundary;
-- no independent outer entry or exit motion.
-
-The Sequence has one explicit terminal point. For activation frames `p1..pN` and terminal `T`:
-
-```text
-p1 < p2 < ... < pN < T
-
-logical member 1 phase [p1, p2)
-logical member 2 phase [p2, p3)
-...
-logical member N phase [pN, T)
-```
-
-Source occurrence order is authoritative. The compiler never sorts by physical time. Missing,
-equal-frame, reversed or out-of-program activations fail the complete Sequence atomically.
-
-Every adjacent pair owns one explicit Handoff. No transition default exists only in an editor or
-Runtime; “apply to all” materializes the same chosen transition on every edge.
-
-### Pair transition
-
-The official initial set remains:
-
-- cut;
-- crossfade;
-- push;
-- wipe;
-- cover;
-- page turn.
-
-One transition owns operator identity, exact duration, typed parameters and a boundary-position
-ratio `r` in `[0,1]`. At logical boundary `p` and duration `d`:
-
-```text
-handoff.start = p - r × d
-handoff.end = p + (1-r) × d
-```
-
-Frame quantization preserves the exact total transition length. Adjacent handoff windows may not
-overlap; invalid duration fails instead of creating an implicit three-source transition.
-
-During a handoff, outgoing and incoming Member surfaces are both owned by the Sequence. Their media
-samples run continuously across their expanded visual spans. The transition changes neither
-ProgramSpace nor upstream Selection/Moment points.
-
-Sequence handoffs may operate only inside their shared Placement Frame. This makes cut, crossfade,
-push, wipe, cover and page turn honest local relationships. An outer handoff against
-`composite_below` is deliberately retired: it would read and mutate unrelated lower Tracks.
-
-The Sequence itself may use an ordinary entry/exit motion. A full-screen fade over a lower Track is
-valid because only the Sequence opacity changes; pushing or page-turning the lower Track is not.
-
-## 11. Audio is explicit and projected separately
-
-Visual material remains muted unless the author explicitly selects source audio from exactly one
-named media layer. This is item-level so a foreground and its self-blurred copy cannot accidentally
-mix the same source twice.
-
-```ts
-type MediaSourceAudioProjection = {
- readonly fromLayer: string;
- readonly gain: number;
-};
-```
-
-The audio projection reuses the selected layer's exact trim, playback mapping and resolved schedule.
-Each Sequence edge separately owns its audio relationship:
-
-- `cut` switches audibility at the logical activation point even when visual surfaces overlap;
-- `crossfade` uses the explicitly authored audio handoff window;
-- visual transitions never silently choose an audio transition.
-
-Entry/exit or handoff SFX are explicit audio Artifact inputs with explicit gain. They are lowered at
-the exact resolved entry, exit or boundary point. They are not package-global filenames, mutable
-sound-library defaults or hidden lane SFX.
-
-The compiler resolves one package-owned Media Program and provides separate deterministic projection
-Operations, rather than one atomic value carrying both terminal Tracks:
-
-```text
-Resolved Media Program ─┬─> project visual ─> VisualTrack
- └─> project audio ─> AudioTrack
-```
-
-Consequently either terminal output may be targeted, replaced or satisfied independently by the
-Run Graph. When both are demanded, they still share the same resolved schedule through an explicit
-graph edge rather than duplicated timing metadata.
-
-A visual-only Item does not demand audio normalization or rendering merely because its source
-container has an audio stream.
-
-## 12. Stacking and flat Tracks
-
-The Media Track itself has no z-index. Every resolved top-level Item or Sequence surface and any
-separately interleavable member/layer emits absolute Present stacking keys.
-
-Frame Paint and tightly coupled media samples may remain in one Present-local element tree. When a
-peer Track must be able to appear between a board, card, icon or layer, the package emits separate
-Presents with separate absolute keys. Track ownership never creates a stacking context.
-
-Temporary ordering inside a Sequence handoff is package-owned and exists only for that relation's
-frames. It does not reserve a global z band.
-
-## 13. Speech Spine's restricted Media projection
-
-Speech Spine should stop maintaining a second ad-hoc media renderer. Its visual projection uses the
-same Media lowering implementation with a deliberately restricted generated Program:
-
-```text
-one normalized speech-bearing Take per Segment; its visual stream is optional
-exact already-established Segment frame span
-explicit authored Spatial Frame
-one foreground layer
-explicit fit and absolute stacking order chosen by the author
-muted visual
-cut between contiguous takes
-no entry, sustain or exit motion
-no Sequence state
-no source-audio projection
-```
-
-The Speech author Surface accepts raw `video=` and expands the same generic inspection and
-normalization graph before assembly, selecting primary moving video plus default audio at the
-Spine's explicit frame rate. Raw `audio=` expands an audio-authoritative normalization graph and
-contributes no visual Present; it never manufactures black media. Prepared `media=` remains a
-direct exact input. Visual Takes inherit the Spine's explicit `visual-frame`,
-`visual-appearance` and `visual-z`, with per-Take overrides for those same limited axes. After assembly, Speech
-audio remains the separate canonical `SpeechAudioBasis -> AudioTrack` projection; the restricted
-visual lowerer itself remains muted and does not rediscover container audio.
-
-This can be a code dependency on focused Media lowering helpers. It does not require exposing a
-that intermediate value as a public graph Type when no external component consumes it. If a
-real graph consumer appears later, the package may expose its own versioned resolved Type; Core
-still does not register Media meaning.
-
-The author imports Speech Spine, whose manifest declares its package dependency. Authors do not
-need to import a second parser merely because the implementation reuses Media lowering.
-
-Ordinary Media Items may consume `during={story.segment.answer}` directly. This is not an implicit
-Selection: Temporal resolves the Segment's already-existing structural start/end anchors. The same
-Item can still consume authored Selection and Moment values when those are the actual intent.
-
-## 14. Legacy audit: retained and retired
-
-Retain:
-
-- full-frame cutaways, split-screen and corner overlays through one Item model;
-- still images, GIF-like animation and videos;
-- independent overlapping media presentations;
-- explicit contain/cover/focal sampling and foreground/self-blur backings;
-- source trim and once/hold/loop/stretch occupancy;
-- entry, exit, continuous outer motion and inner sampling motion;
-- explicit Sequence membership and shared-surface pair transitions;
-- cue-position ratio and continuous visual sampling across transition windows;
-- explicit source audio and transition/edge SFX;
-- stable author identities and deterministic authored order.
-
-Retire:
-
-- a separate B-roll terminal family;
-- a `mode` that changes valid ports or settings;
-- URL-extension guesses for image/video/GIF behavior;
-- track-level z-index and implicit item priority;
-- `auto` playback, `finish`, `freeze`, `phased` and renderer repair;
-- automatic sequencing or `linkNext` inference;
-- one `focalX/focalY` used for both source and destination points;
-- fixed foreground/background fields instead of ordered local layers;
-- one CSS transform slot that lets motion channels overwrite each other;
-- proportional shortening of overlapping entry/exit durations;
-- lane-global SFX and automatically selected audio;
-- outer `composite_below` transitions and every Base FX dependency;
-- one bundled B-roll Product that couples VisualTrack and AudioTrack satisfaction.
-
-The old Deck behavior migrates independently under [`deck-track.md`](./deck-track.md); removing it
-from Media Track does not remove that capability.
-
-## 15. Package and implementation shape
-
-The intended package relationships are:
-
-```text
-@narratage/media-track
-├─ author Surfaces: Track, Item, Sequence, Member, Handoff
-├─ package-owned Program validators
-├─ pure Item window/layer/motion lowering
-├─ pure Sequence schedule + pair transition lowering
-├─ VisualTrack projection
-└─ AudioTrack projection
-
-depends on
-├─ @narratage/media
-├─ @narratage/temporal
-├─ @narratage/spatial
-├─ @narratage/svs
-└─ @narratage/composition
-```
-
-The package may use internal focused files or future implementation-only libraries. It must not
-create one umbrella authoring library shared by Text, Caption, Ranking and Media, and it must not
-require a release of Core when a new media Recipe, motion operator or component package is added.
-
-The former `@narratage/broll` package was used as a regression witness during migration and is now
-retired. Because the project is pre-release, no compatibility package remains. “B-roll” is still a
-useful editorial description for an Item or Sequence, not a terminal Type or author package.
-Source assets and generated files remain untouched.
-
-## 16. Implementation order
-
-Implement only after the shared Temporal and Spatial slices exist:
-
-1. **Implemented:** introduce `@narratage/media-track` with one still-image Item whose BlobArtifact,
- IntrinsicExtent, SpatialFrame, ContentFit and ProgramSpace are separate semantic inputs;
-2. **Implemented:** add timed source trim and every occupancy/alignment policy;
-3. **Implemented:** add ordered Paint/media layers, clipping and the two-frame fit model;
-4. **Implemented:** add lifecycle and sampling motion with the fixed transform stack;
-5. **Implemented:** add explicit Sequence scheduling and pair transitions;
-6. **Implemented:** add separate source-audio and SFX projection;
-7. **Implemented:** route Speech Spine visual projection through the restricted lowerer;
-8. **Implemented:** migrate the checked-in B-roll examples and retire the old package;
-9. **Implemented:** close the package-local acceptance matrix with structural, real-media and
- partitioned-browser evidence. The independent shared Track/Visual IR gates in
- [`track-expressiveness.md`](./track-expressiveness.md) also pass.
-10. **Implemented:** let `video=` lower to the shared Media Pipeline at the connected ProgramSpace
- rate while preserving explicit `media=` as the prepared-media escape hatch.
-
-No step requires a Core, Runtime, queue or Provider change. Surface materialization may use an
-existing explicitly registered Provider capability, but ordinary Item/Sequence lowering is
-deterministic local compilation.
-
-## 17. Acceptance matrix
-
-### Material
-
-- opaque and alpha still images;
-- portrait, landscape and square images;
-- generated/raw video automatically normalized with and without explicitly selected audio;
-- explicitly prepared normalized video connected without a second normalization;
-- animated GIF/WebP normalized to exact frame timing;
-- still and animated Compositable Surfaces;
-- repeated use of one Artifact without duplicated storage or duplicated source audio.
-
-### Item
-
-- Program/absolute windows under `one`, and Selection/Moment windows under `one` and `each`;
-- independent overlapping Items at unrelated absolute stacking keys;
-- every source trim and shorter/equal/longer occupancy case;
-- full Canvas, split, lower card, corner and off-canvas Frames;
-- transparent, solid, gradient, self-blur and alternate-source backing layers;
-- every Content Fit and unequal content/frame focal point;
-- entry, sustain, exit and sampling motion separately and in combination;
-- deterministic composition for overlapping entry/exit durations without hidden retiming.
-
-### Sequence
-
-- image-image, video-video, image-video and alpha-Surface handoffs;
-- strict authored activation order and stable member identity;
-- cut, crossfade, push, wipe, cover and page turn;
-- start/center/end boundary-position ratios;
-- continuous source sampling across handoff overlap;
-- transition duration below, equal to and above the adjacent logical-phase separation when the
- explicit Sequence envelope and no-three-source laws still hold; invalid envelope/overlap fails;
-- missing, repeated, reversed and same-frame activation failure;
-- no implicit three-member overlap and no lower-composite input;
-- group entry/exit operating only on owned pixels.
-
-### Audio and execution
-
-- visual-only source does not demand audio work;
-- exactly one explicitly selected source layer contributes audio;
-- hard and crossfaded Sequence audio handoffs;
-- simultaneous exit/entry SFX mix at one boundary;
-- independent BGM remains unchanged;
-- targeting/replacing visual and audio projections independently;
-- identical Visual Track/Audio Track plans under local and remote execution;
-- arbitrary frame partitioning renders byte-identical deterministic frames.
-
-### Architecture
-
-- installing another Media component changes no Core, Film, Composition or HyperFrames family
- registry;
-- all source, temporal, spatial, sound and label dependencies are graph edges;
-- no Track contains provider, path, Narrative digest or source-family metadata;
-- no operation samples the accumulated lower composite;
-- the Speech Spine restricted projection and ordinary Item use the same lowering laws.
-
-### Executable evidence
-
-The matrix above is closed by executable evidence rather than by package status prose:
-
-- `packages/media-track/test/media-track.test.ts` covers every Item binding, strict occurrence
- expansion, all documented Frames and Content Fits, ordered Paint/sample/backing layers, trim and
- occupancy, lifecycle/sustain/sampling motion, every Handoff operator and ratio, exact source
- audio/SFX projection, malformed Sequence rejection and explicit author-graph edges;
-- `packages/provider-media-local/test/provider.test.ts` runs real FFmpeg inspection and
- normalization for moving video with and without audio, attached pictures, source A/V offsets,
- animated GIF, animated WebP, exact audio rendering and final muxing;
-- `packages/media-execution/test/webp.test.ts` separately proves animated WebP straight-alpha
- blend and dispose behavior used by both local and Lambda execution;
-- `packages/provider-media-aws-lambda/test/provider.test.ts` proves that remote execution receives
- the same exact `AudioProgramPlan`; both local and Lambda media environments execute
- `@narratage/media-execution` rather than owning alternate media semantics;
-- `packages/speech-spine/test/surface.test.ts` and `packages/speech-basis/test/product.test.ts`
- prove the restricted Speech projection, one shared generated Product and independent visual/audio
- Candidate satisfaction;
-- `packages/hyperframes/test/browser-visual.test.ts` renders opaque images, a straight-alpha
- Surface, two-frame fitting, local motion and a Sequence handoff through real Chromium, then
- compares every decoded RGBA frame across different worker partitions;
-- package manifests contain no Media family registry, while the public value types contain no
- hidden lineage or Provider metadata.
-
-The local package is therefore complete as a pre-release Media implementation over the frozen
-repository-internal `svml.visual-track@1` / `svml.visual-ir@1` waist. This does not publish or
-separately freeze the Media author Surface.
diff --git a/docs/guide/spec/ranking-track.md b/docs/guide/spec/ranking-track.md
deleted file mode 100644
index 337e64b7..00000000
--- a/docs/guide/spec/ranking-track.md
+++ /dev/null
@@ -1,310 +0,0 @@
----
-title: SVML Ranking Track Migration
-description: implemented executable authority for the official Ranking migration. All four author.
----
-
-# SVML Ranking Track Migration
-
-Status: implemented executable authority for the official Ranking migration. All four author
-components, their shared triggered schedule, optional peer AudioTrack lowering, author Surfaces and
-real-browser evidence execute in `@narratage/ranking`. It is not a compatibility promise for the
-historical node format or a frozen public ABI.
-
-## 1. What Ranking means
-
-Ranking is a video-domain author component that presents one ordered collection as a progressive
-visual state. It is not a Core concept, a generic Track mode, a Provider or a renderer feature.
-
-The old system placed four materially different behaviors behind one `ranking_track` node and a
-`rankingType` switch. The new system uses one physical package, `@narratage/ranking@1`, but exposes
-four independent author components:
-
-- `ranking:TierBoard` — place revealed items into named tier rows;
-- `ranking:Column` — show one current item on a stage, then settle it into a numbered column;
-- `ranking:TopThree` — progressively reveal up to three items and accent the current item;
-- `ranking:TypewriterList` — type ordered text rows onto a persistent paper/list surface.
-
-One package is the right unit for shared validation, schedule construction, layout utilities and
-paint primitives. Four components are the right unit for author meaning. There is no public
-`rankingType`, mode-selection input, dynamic-port registry or Runtime renderer registry.
-
-A third party can publish another Ranking package without registering a type in Core or modifying
-this package. Installing and importing that package contributes its own Surface, Types, validators
-and fragment compiler through the ordinary module mechanism.
-
-## 2. Evidence from the old implementation
-
-The reviewed old-project inventory contains 21 migrated Ranking nodes:
-
-| Historical variant | Reviewed uses | Item counts seen |
-|---|---:|---|
-| Tier | 10 | 5–6 |
-| Typewriter | 5 | 5 |
-| Column | 3 | 4–5 |
-| Top Three | 3 | 3 |
-
-The inventory proves that all four meanings are real, but not that their old representation should
-survive. The old implementation had these structural problems:
-
-1. one UI/node identity changed meaning by switching a mode and disconnecting its dynamic ports;
-2. the same item `start/end` pair meant visibility, animation phase, successor handoff or nothing,
- depending on the renderer;
-3. Top Three ignored item end times, while Typewriter used their union only to decide whether the
- paper was mounted;
-4. renderers inspected all sibling segments, sorted or counted them, and manufactured group state
- after the generic Track Program had already lost that meaning;
-5. one global z-index prevented the board and individual icons from interleaving with peer Tracks;
-6. raw URLs, environment font names, Remotion component ids and three mirrored registries leaked
- environment/renderer details into the component;
-7. short windows could silently suppress motion sound instead of fitting visual and sound phases to
- one shared finite schedule.
-
-Migration preserves the useful visual capabilities, not these accidental semantics.
-
-## 3. Author-facing shape
-
-The executable author Surface is:
-
-```svml
-<import as="ranking" from="@narratage/ranking@1"/>
-<import as="fonts" from="@narratage/fonts-open@1"/>
-<import as="studio" source="./studio.svs"/>
-
-<fonts:Stack id="ranking-fonts" family="inter" weight="700" style="normal"/>
-
-<ranking:ColumnStyle id="column-style"
- recipe={studio.ranking.column}
- font={ranking-fonts}/>
-
-<ranking:Column id="tools"
- space={speech.space}
- map={timing.map}
- during={story.selection.ranking-board}
- triggers={story.moment.next-rank}
- terminal={story.moment.ranking-complete}
- style={column-style}>
- <ranking:ColumnItem label="Fourth" icon={fourth.image}/>
- <ranking:ColumnItem label="Third" icon={third.image}/>
- <ranking:ColumnItem label="Second" icon={second.image}/>
- <ranking:ColumnItem label="First" icon={first.image}/>
-</ranking:Column>
-```
-
-`during` supplies the explicit outer window. `triggers` supplies an ordered Moment occurrence set.
-`terminal` is the explicit end of the final active stage; it may precede the outer end so the fully
-settled result remains visible. The component requires the number of trigger occurrences to equal
-the number of child items and zips them in authored order.
-
-The outer Selection and terminal Moment each use strict `one` semantics. The trigger Moment uses
-strict `each` semantics. Zero or multiple outer/terminal occurrences, zero triggers and any
-trigger/item cardinality mismatch are compile errors; the component never selects a convenient
-occurrence.
-
-This positional zip is declared by these four components. It is not generic Core behavior. An
-alternative third-party Surface may give each item its own Moment edge and lower to the same
-package facts.
-
-Style declarations remain separate from use. SVS supplies named generic Recipe data; each
-variant-specific `*Style` component owns the allowed keys and compiles the Recipe plus explicit font
-edges into a typed package-owned Style. Semantic item values such as tier row, label, icon and entry
-behavior are not paint properties and do not belong in SVS.
-
-## 4. The group sees siblings; rendered Tracks do not
-
-A progressive ranking cannot be compiled by treating every child as an isolated Track item. Row
-placement, successor handoff, active-item state and the settled prefix depend on the ordered group.
-That does not mean items inspect one another at render time.
-
-The author component first compiles the complete child list and trigger set into one package-private
-schedule:
-
-```ts
-type RankingSchedule = {
- readonly outer: FrameSpan;
- readonly terminalFrame: number;
- readonly entries: readonly {
- readonly itemId: string;
- readonly triggerOccurrenceId: string;
- readonly triggerFrame: number;
- readonly stage: FrameSpan;
- readonly cumulative: FrameSpan;
- }[];
-};
-```
-
-This is an explicit Graph value between deterministic Operations, not metadata smuggled through an
-image, Track or media Artifact. It is owned and versioned by the Ranking package and is not added to
-Core, Composition or the shared Track waist.
-
-```text
-Moment set + SemanticMap + ProgramSpace + outer/terminal projection + ordered item ids
- │
- ▼
- RankingSchedule
- │ │
- Style + content ───┘ └── SFX clips + gains
- │ │
- ▼ ▼
- VisualTrack AudioTrack
-```
-
-The schedule Operation preserves source order and stable occurrence identities. It never sorts by
-physical time. After frame quantization it requires:
-
-```text
-outer.start <= p1 < p2 < ... < pN < terminal <= outer.end
-```
-
-Every invalid or equal-frame sequence fails atomically. The visual and optional audio lowerers
-consume the same schedule, so motion and sound cannot drift while remaining independently
-demandable branches of the Run Graph.
-
-## 5. Common phase model
-
-For outer window `[B, O)`, triggers `p1 ... pN` and terminal `T`, entry `i` receives:
-
-```text
-active stage [p_i, p_(i+1)) or [p_N, T)
-cumulative life [p_i, O)
-settled life [stage.end, O)
-```
-
-`[B, p1)` may show an explicitly authored empty board, but has no implicit active item. `[T, O)` is
-the optional fully-settled suffix with no active item. These facts are package-owned sibling
-sequencing built on the shared temporal projection algebra in
-[`track-authoring.md`](./track-authoring.md); they are not a new staging contract.
-
-### TierBoard
-
-- the board occupies `[B, O)`;
-- every item is visible cumulatively from its trigger to `O`;
-- `entry="direct"` animates directly into its final tier cell after the trigger;
-- `entry="stage"` appears at the shared stage point, remains there for its active stage, moves into
- the final cell before the stage ends and then stays settled;
-- the item owns a required image and tier-row id; source order within a row determines its cell;
-- impossible appear/move durations are compile errors, never silently shortened or omitted.
-
-### Column
-
-- the numbered board occupies `[B, O)`;
-- exactly one item occupies the large stage during each active stage;
-- when its stage ends, that item becomes part of the settled column until `O`;
-- label is required, image is optional, and rank is the authored item order;
-- overlapping current items are structurally impossible after a valid schedule.
-
-### TopThree
-
-- at most three items are accepted;
-- each item is revealed cumulatively from its trigger to `O`;
-- only the item for the current active stage receives the active ring/accent motion;
-- `[T, O)` shows all revealed items without an active accent;
-- label is required, image is optional, and slot is the authored item order.
-
-### TypewriterList
-
-- the paper occupies `[B, O)` and blank rows may be structurally reserved before their trigger;
-- row `i` begins typing at `p_i` and remains visible until `O`;
-- typing duration is derived deterministically from Unicode grapheme count and the Style's speed;
-- typing and an optional winner mark must fit inside the item's active stage or compilation fails;
-- title, optional explicit emphasis span and row strings are author content, not LLM fields.
-
-Typewriter is retained in the Ranking package because its author meaning is an ordered progressive
-ranking, despite its heavy use of text layout. It may share low-level exact-font/layout utilities
-with the future Text implementation, but it must not depend on or masquerade as a generic Text
-Track Program.
-
-## 6. Spatial layout and stacking
-
-Each component owns one outer placement box and its internal layout algorithm. A Tier cell, Column
-row or Top Three slot is not a generic per-item spatial locator. Allowing arbitrary spatial inputs
-on those children would destroy the component's meaning.
-
-Placement and stacking remain independent:
-
-- the outer box determines the coordinate system for board, stage and items;
-- the Style resolves an absolute board stack order and defaults for stage/items;
-- every child may explicitly override its own absolute item stack order;
-- board, stage decoration and each item lower to separate `VisualPresent` values where independent
- interleaving is required;
-- one resulting `VisualTrack` is not a stacking context, so a peer Track Present may appear between
- the board and any icon.
-
-No package must emit multiple Tracks merely to obtain multiple z positions. Stable `tieBreak`
-values derive from authored component/item identity, never evaluation order.
-
-All normalized geometry is relative to the explicit outer box. Required variant dimensions include:
-
-- Tier: row definitions/colors, board/label widths, stage point, cell gaps, icon fit/radius and
- appear/move durations;
-- Column: board box, row height/gap, stage point/size, cell/icon geometry, rank palette and
- appear/move durations;
-- Top Three: center/width, label line, icon size, ring width/palette and reveal/accent motion;
-- Typewriter: paper theme/paint, padding, row gap, title/item typography, rotation, typing speed and
- winner-marker paint/motion.
-
-Paint, geometry and motion parameters are validated by the owning `*Style` component. Unknown
-Recipe properties fail. There is no arbitrary CSS escape hatch.
-
-## 7. Media, fonts and sound
-
-Images and sounds enter through explicit `BlobArtifact` edges. Fonts enter through explicit,
-content-addressed font stack edges. Ranking never stores a URL, asks the Runtime to choose a font,
-or names a renderer component.
-
-Optional sound effects are ordinary authored inputs:
-
-- appear sound starts at the item's trigger;
-- move sound is derived from the same scheduled move phase used by the visual lowerer;
-- gain and clip choice live in the Style/explicit sound inputs;
-- configured sound produces a peer `AudioTrack`; no sound produces no demanded audio branch;
-- sound playback never enters Hyperframes or a visual Track.
-
-Ranking itself is deterministic and has no Need, Provider, queue, credential or environment
-adapter. If an input image is generated, that upstream generation remains another Graph component.
-
-## 8. Run-Graph behavior
-
-The author graph exports ordinary logical `visual` and, when authored, `audio` results. Targets,
-Candidates and replacement are governed by the same Run Graph laws as every other component.
-Ranking adds no cache, pin or preview primitive.
-
-Replacing the visual result does not mutate the audio result. Replacing both outputs with one
-multi-output Candidate is also possible through ordinary explicit output mapping. Once every
-demanded original output is satisfied by selected Candidates, demand analysis prunes the original
-Ranking branch and its temporal/media ancestors. No special Ranking execution rule is required.
-
-## 9. Migration order and acceptance
-
-Implemented in this order:
-
-1. **Implemented:** the shared temporal projection and triggered-schedule laws;
-2. **Implemented:** `@narratage/ranking@1` with four separate Programs/Styles and one private schedule utility;
-3. **Implemented:** TierBoard and Column, which together prove cumulative and exclusive stage
- semantics plus independent stacking;
-4. **Implemented:** TopThree;
-5. **Implemented:** TypewriterList using exact font/layout primitives;
-6. **Implemented:** optional AudioTrack lowering from the same schedule;
-7. **Implemented:** a real-browser progressive-state witness and one author-Surface graph check per component.
-
-Acceptance requires:
-
-- trigger/item cardinality, stable identities and all boundary failures are tested;
-- `T < O` visibly proves a final settled suffix;
-- short stages deterministically fit their local animation and sound phases without changing stage
- boundaries; a very short Typewriter stage may reveal several graphemes on the same frame;
-- board and per-item Presents interleave with a peer Track at unrelated absolute z positions;
-- visual and sound event frames agree exactly;
-- missing optional images work only for variants that declare them optional;
-- fonts and media are content-addressed dependencies;
-- each component can revise its own Program without changing Core, Runtime, Composition,
- Hyperframes or another Ranking component;
-- installing a fifth component package requires no central family registry change.
-
-The migration intentionally does not preserve historical node JSON, mode switching, old dynamic
-ports, renderer ids or Remotion components. Historical projects are evidence and visual references,
-not a compatibility obligation for an unreleased language.
-
-The package test matrix covers every acceptance item above. The opt-in Chromium witness evaluates
-all four components with one worker and partitioned workers, proves visible trigger progression and
-the final settled suffix, and bounds only microscopic cross-process edge-antialiasing noise. Logical
-frame state, geometry, stacking, event frames and authored graph edges remain exact.
diff --git a/docs/guide/spec/screen-overlay.md b/docs/guide/spec/screen-overlay.md
deleted file mode 100644
index e7c9dd32..00000000
--- a/docs/guide/spec/screen-overlay.md
+++ /dev/null
@@ -1,247 +0,0 @@
----
-title: SVML Screen Overlay Authoring
-description: implemented pre-release contract for the optional official Screen Overlay package. Its.
----
-
-# SVML Screen Overlay Authoring
-
-Status: implemented pre-release contract for the optional official Screen Overlay package. Its
-self-described Surface, eleven typed components and sequential/parallel browser evidence execute;
-it is not yet a frozen public ABI.
-
-## Purpose
-
-`@narratage/screen-overlay` contributes self-contained, usually full-canvas pixels over an authored
-time window. A flash, vignette, scan-line veil or deterministic light leak is an ordinary
-`VisualTrack`, exactly like Text, Caption, Media or Ranking.
-
-It is not an adjustment layer, a privileged post-composition hook, a transition engine or a way to
-read the pixels of lower Tracks.
-
-```text
-Moment / Selection / Program ─> shared temporal projection ─┐
-CanvasSpace ────────────────────────────────────────────────┼─> Overlay Program
-effect parameters / explicit owned assets ─────────────────┘ │
- ▼
- Visual IR elements or owned Surface
- │
- ▼
- VisualTrack
-```
-
-Core, Runtime and Composition do not know Screen Overlay. Composition only validates and orders its
-self-contained Presents.
-
-## 1. The self-contained-pixel law
-
-Every overlay item must be renderable from:
-
-- its own authored parameters;
-- the explicitly connected CanvasSpace and ProgramSpace;
-- zero or more explicit assets owned by that item;
-- its deterministic local frame number.
-
-It must not sample:
-
-- the accumulated Composition below it;
-- a sibling Track;
-- a hidden Base Track;
-- the previous or next scene;
-- a browser backdrop or environment texture.
-
-Normal source-over alpha compositing is allowed: that is how every alpha-bearing Track reaches the
-canvas. Blend modes or filters whose result depends on unknown lower pixels are not self-contained.
-
-This law means the old source-free `backdrop-filter` implementations of Gaussian Blur, Color
-Adjust and Zoom Blur are not Screen Overlays. To preserve those effects honestly, a media-effect
-component must receive the exact media or compositable Surface to transform through a graph edge,
-then output its own transformed material or Track. If a future component explicitly rasterizes a
-selected group of Tracks into a Surface, consuming that Surface is also explicit; Screen Overlay
-never gains a secret “current frame” input.
-
-## 2. Author package, not operator registry
-
-The official package may expose several readable author components under one namespace:
-
-```xml
-<screen:Track id="screen-paint" canvas={vertical} space={speech.space}>
- <screen:Flash at={story.moment.hit} map={timing.map} for="240ms" z="90"
- color="#ffffff" intensity="0.9" attack="2" hold="3" decay="5"/>
- <screen:Vignette during="program" z="20"
- center-x="0.5" center-y="0.5" radius-x="0.82" radius-y="0.68"
- softness="0.3" color="#000000" opacity="0.28"/>
- <screen:Grain during="program" z="70"
- amount="0.12" size="3" chroma="monochrome" motion-rate="0.5" seed="23"/>
-</screen:Track>
-```
-
-This is illustrative Surface syntax. The compiled Program contains explicit component identity,
-parameters and temporal binding; it is not one stringly typed `effect="..."` switch owned by Core.
-
-One package is convenient for the small official collection, but it is not a central registry. A
-third party may publish another self-described package that lowers a new component to the same
-Visual Track waist. Adding an overlay must not require changing Core, Film, Composition or
-HyperFrames when existing Visual IR or a typed Surface can express it.
-
-Appearance defaults and named combinations belong in `.svs` recipes. The package owns parameter
-validation and deterministic lowering. Runtime Profiles and Providers do not choose a visual
-effect.
-
-## 3. Program axes
-
-Every overlay item has four independent author concerns:
-
-```ts
-type ScreenOverlayItemSpec = {
- readonly id: string;
- readonly content: ScreenOverlayComponent;
- readonly projection: TemporalWindowProjection;
- readonly expansion: OccurrenceExpansion;
- readonly stackingOrder: number;
-};
-```
-
-- `content` identifies the actual visual component and its typed parameters;
-- component-owned envelope and local-motion parameters remain inside that typed component rather
- than a universal presentation bag;
-- `projection` and `expansion` use the shared Selection/Moment/Program algebra;
-- `stackingOrder` resolves with stable authored identity to an ordinary absolute Present key.
-
-Canvas geometry is not an author-selected Spatial Frame here. A Screen Overlay intentionally owns
-the complete connected CanvasSpace. A partial-frame light, image, card or blur is a normal Media,
-Text or component Track and uses [`spatial-layout.md`](./spatial-layout.md).
-
-Items are independent by default. Overlap is valid and produces multiple ordinary Presents. There
-is no singleton Screen Overlay lane and no global merge step.
-
-## 4. Timing and local envelopes
-
-Selection, Moment, Program and absolute timing lower through
-[`track-authoring.md`](./track-authoring.md). Screen Overlay does not have `from`, `until`, `manual`
-or phrase-matching timing laws of its own.
-
-The projected window answers only where the item exists. Its within-window envelope is component
-presentation. For example, a Flash may define attack, peak and decay; a Directional Matte may move
-an edge from one side to another; Grain may remain steady. The old universal triangle that peaked
-at 30% of every effect's duration is retired as hidden presentation policy.
-
-Recipe defaults may supply an envelope, but the resolved Program must contain the exact parameters.
-An effect cannot expand its projected window to fit an animation. A longer authored envelope is
-simply clipped by that window; it is not rejected or secretly sped up. Invalid or zero-length
-windows still fail through the common temporal validator.
-
-## 5. Portable official component set
-
-The useful old effects divide into two honest lowering groups.
-
-### Direct Visual IR candidates
-
-These can normally lower to full-canvas boxes, gradients and local keyframes:
-
-- `Flash` — solid color with an explicit opacity envelope;
-- `ColorWash` — source-over color veil; no hidden lower-pixel blend mode;
-- `Vignette` — authored center, radius, softness, roundness, color and opacity;
-- `ScanLines` — spacing, thickness, angle, opacity and local motion;
-- `DirectionalMatte` — angle, coverage, feather, color, opacity and progress;
-- `WhipVeil` — direction, width, softness, travel and opacity;
-- `GlitchVeil` — deterministic owned color bars/washes, not displacement of lower pixels.
-
-### Owned Surface candidates
-
-These may be easier or more exact as deterministic alpha-bearing `CompositableSurface` values, but
-the current official implementations fit exactly in the closed code-free Visual IR:
-
-- `Grain` — amount, grain size, monochrome/color, motion rate and required seed;
-- `LightLeak` — colors, angle, softness, travel, intensity and seed when stochastic;
-- `Bokeh` — amount, size range, color/warmth, drift and required seed;
-- `TVStatic` — amount, noise size, scan-line contribution, motion rate and required seed.
-
-Choosing Visual IR or Surface is a lowerer implementation decision only when both produce the exact
-declared pixels. It cannot depend on which Runtime happens to execute the Build. A materializing
-implementation may issue an ordinary Provider Need, but the Need returns a typed owned Surface and
-does not receive the lower composite.
-
-All stochastic behavior requires an explicit resolved seed. `Math.random()`, wall time, GPU noise
-or renderer-dependent entropy is invalid.
-
-## 6. Explicitly excluded effects
-
-The following old names are excluded from the source-free package:
-
-| Old effect | Why excluded | Honest replacement |
-|---|---|---|
-| `gaussian_blur` | must read pixels being blurred | media/surface filter with explicit input |
-| `color_adjust` | exposure/contrast/saturation/hue alter input pixels | media/surface color transform with explicit input |
-| `zoom_blur` | a real radial blur requires the visual being sampled | media/surface effect with explicit input |
-| screen/multiply/overlay blend wash | result depends on lower pixels | source-over alpha approximation, or explicit-input materialization |
-
-This does not delete the visual result. It rejects the old privilege by which a supposedly ordinary
-Track could mutate everything beneath it.
-
-Screen Overlay also cannot:
-
-- change Program duration or playback rate;
-- own the cut between adjacent media items;
-- transition two sibling Tracks;
-- reserve a topmost z-index;
-- hide a sound effect inside a visual item.
-
-A visually masked hard cut remains two independent facts: the Media package owns the cut and the
-Screen Overlay happens to cover some of the same frames. A combined authored Fragment may
-produce one Visual Track and one Audio Track from the same Moment, but both outputs remain explicit
-peer edges into Film.
-
-## 7. Stacking
-
-The old fixed `z_index: 200` rule is retired. “Screen” describes full-canvas geometry, not a magical
-render phase. Every Present receives an explicit absolute stacking intent and Composition flattens
-it with all other Presents.
-
-An author can intentionally put a vignette below captions, a flash above captions or grain between
-two graphic layers. Package recipes may provide convenient stacking defaults, but Film and
-Composition never switch on the package family and never force it to the top.
-
-If one overlay package emits multiple Presents at different stacking positions, peer Tracks may
-interleave between them exactly as required by the flat Track law.
-
-## 8. Legacy migration audit
-
-Retain:
-
-- short Moment hits and long Program overlays;
-- Selection, Moment and absolute temporal placement;
-- full-canvas flash, veil, vignette, light leak, grain, scan lines, bokeh and static visuals;
-- deterministic effect-specific parameters and local envelopes;
-- several simultaneous effects;
-- optional explicit owned textures or materialized alpha Surfaces.
-
-Retire:
-
-- singleton slot semantics;
-- fixed topmost z-index and special post-caption render phase;
-- one central string operator registry as system truth;
-- the universal 30%-peak envelope;
-- hidden `backdrop-filter` access;
-- `mix-blend-mode` across Tracks;
-- pretending screen hits are scene transitions;
-- hidden lane-level enter/exit SFX;
-- any Base FX counterpart or placeholder.
-
-## 9. Lowering and acceptance
-
-Implementation follows the shared Temporal and existing Visual Track contracts:
-
-1. **Implemented:** the self-described `@narratage/screen-overlay` author package;
-2. **Implemented:** eleven separate typed author components rather than one unvalidated bag;
-3. **Implemented:** exact code-free `svml.visual-ir@1` lowering; the generic typed owned-Surface
- escape remains available when a future component cannot be represented exactly;
-4. **Implemented:** connected CanvasSpace and ProgramSpace, never Film globals;
-5. **Implemented:** ordinary absolute-stack Presents;
-6. **Implemented:** deterministic structural/frame tests for every component and envelope;
-7. **Implemented:** identical pixels across one-worker and three-worker real Chromium rendering;
-8. **Implemented:** multiple Overlay Tracks interleave with peer Tracks by stacking key;
-9. **Implemented:** negative tests rejecting any lower-composite, sibling-Track, backdrop-filter or hidden audio
- dependency.
-
-The design fails if installing a new overlay requires a Core branch, a Composition family, a
-HyperFrames component switch or an implicit snapshot of already composed pixels.
diff --git a/docs/guide/spec/script-surface.md b/docs/guide/spec/script-surface.md
deleted file mode 100644
index d1e6fbcd..00000000
--- a/docs/guide/spec/script-surface.md
+++ /dev/null
@@ -1,696 +0,0 @@
----
-title: SVML Script Surface
-description: "Normative contract for the SVML script body: Segment, Role Cue, Dual Text, Selection, Moment and Slot."
----
-
-# SVML Script Surface
-
-> **Draft; not frozen.**
->
-> SVML has not been publicly released, so there is no legacy language version to remain
-> compatible with. This document directly defines the first public target, Script Surface
-> v1; the early prototypes in this repository are not the normative contract. An
-> implementation may claim SVML v1 support only after it passes the parser, projection,
-> source-map and temporal golden fixtures in this document.
-
-SVML is a semantic video source language, not a pixel rendering format. This document
-defines only its most frequent and most readability-critical script region, `<script>`; the
-outer document structure for canvas, nodes, styles, generation parameters and the other
-Program regions is designed separately. After v1 ships, new semantics should first be
-composed out of the constructs already in this document; only when the body syntax genuinely
-must change should a new version be published, and early prototype behaviour must never be
-used to weaken the current v1.
-
-## One-page contract
-
-Apart from natural language, the body has only six semantic constructs:
-
-| Construct | Owns only | Does not own |
-|---|---|---|
-| `Segment` | Ordered media/spoken blocks and independent structural endpoints | Automatic Selection, duration, silence |
-| `Role Cue` | The "who said what" text prefix in the `dialogue` projection | speaker entities, voice timbre, field routing |
-| `Dual Text` | Display text that differs from the actual pronunciation | TTS vendor parameters, styling, motion |
-| `Selection` | One or more explicitly closed semantic time intervals | Consumer declarations, absolute seconds |
-| `Moment` | One or more explicitly complete semantic instants | Half intervals, durations |
-| `Slot` | Injecting runtime text values | Injecting SVML syntax or arbitrary AST |
-
-Comments are trivia, not a seventh semantic construct. Every caption style, B-roll, Ranking,
-Deck, sound effect, silence, generation prompt, node port and consumption policy references a
-Selection or Moment from outside `<script>`. A consumer does not belong to any one Segment;
-one consumer may have several ports, and each port may take one or more Selections/Moments. A
-Selection may cross Segments or consist of disconnected intervals, and a Moment may likewise
-have several occurrences.
-
-Script compilation MUST first produce a Narrative IR that preserves source mapping, and then
-project the three texts from that same IR. Producing three unrelated strings first and then
-guessing the correspondence between them is forbidden.
-
-| Projection | Content | Typical consumers |
-|---|---|---|
-| `dialogue` | Role Cue + actual pronunciation | Dialogue/video generation such as Seedance |
-| `speech` | Actual pronunciation only | Estimate, TTS, Speech Reference, alignment |
-| `caption` | Display text only | Captions and text display |
-
-## Worked example
-
-```svml
-<script>
- @whole
-
- <hook>
- <A> I just @laugh @pop! <lmao | laughed my @punch ass out @/punch> @/laugh.
- </hook>
-
- @silence
- <pause/>
- @/silence
-
- <close>
- <B> Meet <${product} | ${product_pronunciation}>.
- </close>
-
- @/whole~
-</script>
-```
-
-If `product = "SVML"` and `product_pronunciation = "ess vee em ell"`, the three text
-projections are:
-
-```text
-dialogue
-A: I just laughed my ass out.
-B: Meet ess vee em ell.
-
-speech
-I just laughed my ass out.
-Meet ess vee em ell.
-
-caption
-I just lmao.
-Meet SVML.
-```
-
-`silence` is only the name of a Selection; it does not automatically mute the audio. It
-completely selects the two independent structural endpoints of the empty Segment; actual
-silence or footage behaviour is decided by the outer Program. `pop` is a Moment with the
-default right affinity, resolved to the onset of `laughed`; it likewise enters no text
-projection.
-
-## 1. Document and Segment
-
-In the official Markup Frontend, one `<script>` Surface instance produces one Narrative; the
-`<script>` shell is dispatched by Text according to the post-import Surface Registry, and the
-Script Parser receives only its raw body. A Script body has only two Segment forms, the named
-block and the named empty block:
-
-```svml
-<intro>
- Hello.
-</intro>
-
-<pause/>
-```
-
-Normative rules:
-
-- A Script contains one or more Segments in source order.
-- A Segment's tag name is its id; the id is unique within the current Script, and Segments
- MUST NOT nest.
-- A Segment opening accepts no attributes; `script` is the reserved name of the outer raw
- Surface and MUST NOT be used as a Segment id.
-- Natural language may appear only inside a Segment; between Segments only whitespace,
- Selection/Moment markers and comments are allowed.
-- An empty Segment uses the self-closing form. It has structural start/end cut points, but no
- lexical anchor.
-- An empty Segment implies neither silence nor any default duration. Its physical duration
- comes from media, from a generation result, or from a Program outside `<script>`.
-- A Segment does not automatically declare a Selection of the same name. Selecting one, three
- or any number of Segments MUST be written as an explicit closed Selection.
-- Every Segment independently owns two stable identities, `start` and `end`; adjacent
- Segments do not share structural endpoint identity. The two identities may land on the same
- ProgramPoint at a hard cut, or on different points because of overlap or gap, but they MUST
- NOT be merged into one identity.
-
-For example, selecting three consecutive Segments:
-
-```svml
-@chapter
-<one>
- One.
-</one>
-<pause/>
-<two>
- Two.
-</two>
-@/chapter
-```
-
-## 2. Role Cue
-
-A Role Cue appears only after the parser has entered some Segment body, and it opens a spoken
-turn. Its recognition is determined by parse state, independent of logical line starts and
-physical newlines:
-
-```svml
-<dialogue>
- <A> What time
- is it?
- <B> It’s 8:30.
-</dialogue>
-```
-
-The following compact form has exactly the same Narrative semantics:
-
-```svml
-<dialogue><A> What time is it?<B> It’s 8:30.</dialogue>
-```
-
-In the `dialogue` projection the canonical serializer emits it as:
-
-```text
-A: What time is it?
-B: It’s 8:30.
-```
-
-In the `speech` and `caption` projections, both `<A>` and `<B>` are removed. A Role Cue is not
-a speaker data model: it creates no character entity, selects no voice, and does not implicitly
-group captions. An outer Program may explicitly write `role="A"` to query those spoken turns.
-The Caption Program lowers such a query directly into a `CaptionDisplayWordSubset`; it is not a
-temporal Selection and does not fabricate a Selection Record. `<A>` on its own triggers no
-styling, character, voice or footage behaviour.
-
-To remove ambiguity:
-
-- In Segment body state, a bare angle-bracket construct that contains no unescaped `|`, has no
- attribute and no `/`, and satisfies the Role label character constraints is recognized as a
- Role Cue; it MUST be followed by non-empty spoken content before the next Role Cue or the end
- of the Segment.
-- A colon in the body is always body text; the parser never guesses a speaker from `A:`.
-- A single non-empty Segment either has no Role Cue at all, or its first spoken atom MUST be
- opened by a Role Cue. Temporal markers, comments and layout whitespace do not count as spoken
- atoms; a Segment that has already begun with cue-less body text MUST NOT switch to cued mode
- partway through.
-- A Role Cue's turn continues until the next Role Cue or the end of the Segment. Physical
- newlines inside a turn are layout whitespace only: they do not end the turn, and they do not
- require the Role Cue to be repeated.
-- A Role Cue is not a container tag and has no close syntax; forms such as `</A>` and `</B>`
- MUST fail as unknown angle-bracket constructs, and the formatter never emits them.
-- A Role label is readable text of 1–32 Unicode characters after NFC; it may consist of Unicode
- Letter, Mark, Number, interior spaces, `_`, `-` and `.`, and MUST NOT have leading or trailing
- whitespace. Quotes, `=`, `/`, newlines, `<`, `>`, `|` and `:` are all illegal, so an unknown
- structural tag carrying attributes is never swallowed as a Role Cue.
-- In the `dialogue` projection, the canonical serializer emits a Role Cue boundary as one
- newline plus a `label: ` prefix; in `speech` and `caption` only the cue is removed, and the
- necessary separation between adjacent atoms is preserved. Re-wrapping the source MUST NOT
- change the semantic content of the Narrative IR or of any text projection; the only things
- allowed to change are the source range and the source hash.
-
-## 3. Dual Text
-
-Dual Text writes out the display text and the actual pronunciation at the same time:
-
-```svml
-<lmao | laughed my ass out>
-```
-
-The left side enters `caption`; the right side enters `dialogue` and `speech`. Ordinary body
-text is shorthand for `<text | text>` and needs no markup.
-
-Dual Text is an atomic, non-nestable correspondence unit, not a simple string substitution:
-
-- Either side may be an arbitrary N:M text mapping; the word counts need not be equal.
-- After layout whitespace is removed, the right side MUST be non-empty; text that is only
- displayed and never spoken does not belong to Script and should be expressed by an outer
- Text/Deck/Caption Program.
-- The left side may be empty: for example `< | um>` means the filler is actually spoken but
- not displayed in captions.
-- Slots may be used on both sides; Selections/Moments may appear only on the speech side, and
- only at its lexical token boundaries.
-- The entire display side inherits the time span of the entire speech side. If a display word
- needs independent timing, selection or styling, the author MUST split it into several Dual
- Text atoms.
-- An ordinary temporal consumer may select part of the tokens on a Dual Text speech side; but
- if a caption ownership/style consumer covers only part of the speech span of a Dual Text
- atom, compilation MUST fail with `partial dual atom` rather than guess how to cut the
- left-side display text.
-
-For example, a legal fine-grained temporal selection:
-
-```svml
-<lmao | laughed my @middle ass out @/middle>
-```
-
-`middle` may be used by B-roll or a sound effect; if it is used as a local caption replacement
-region for that `lmao`, it MUST be an error. The author can provide an explicit mapping by
-splitting the atom.
-
-Dual Text expresses only "what was actually said". IPA, SSML, stress, rate, language, emotion,
-timbre and vendor-specific pronunciation dictionaries are not a fourth body text layer; they
-should be expressed by an outer generation/speech Program that references a Selection or a
-lexicon resource.
-
-## 4. Selection
-
-A Selection is a semantic interval declaration that the author closes explicitly:
-
-```svml
-@id ... @/id
-~@id ... @/id~
-```
-
-Every start marker MUST have an end marker with the same id. Script's public value emits, for
-every occurrence, an already-resolved `startAnchorId` and `endAnchorId`; there is no start-only
-object, no consumer-side default tail completion, and no "half the information was transmitted"
-state. Left/right affinity, token indices and source ranges belong to the Parser/Source Map and
-do not leak into the public Selection. A downstream that needs physical time MUST project those
-two identities together with an explicitly connected `CompleteSemanticMap` into an interval.
-
-"Closed" means both source endpoints are explicitly declared. The physical runtime uses half-open
-`[startFrame, endFrameExclusive)` Frame Spans.
-
-### 4.1 Left and right affinity
-
-The marker itself is zero-width and enters no text projection. It sits between two optional
-semantic boundaries, one on each side:
-
-| Endpoint form | Affinity | Typical meaning |
-|---|---|---|
-| `@id` | Right | Start at the word start on the right or the structural endpoint on the right |
-| `~@id` | Left | Expand outward to the word end on the left or the structural endpoint on the left |
-| `@/id` | Left | End at the word end on the left or the structural endpoint on the left |
-| `@/id~` | Right | Expand outward to the word start on the right or the structural endpoint on the right |
-
-Examples:
-
-```svml
-before @x hello @/x after
-```
-
-`x` runs by default from the word start of `hello` to the word end of `hello`.
-
-```svml
-before ~@x hello @/x~ after
-```
-
-`x` expands outward on both sides, taking the nearest optional boundary to the left of `~@x`
-and to the right of `@/x~` respectively.
-
-```svml
-@pause
-<empty/>
-@/pause
-```
-
-The start of `pause` attaches to the start cut point of the empty Segment, and its end to that
-Segment's end cut point. A Selection may lie entirely outside a Segment, and may also span any
-number of Segments.
-
-### 4.2 Closure, crossing and disconnection
-
-The same id may appear in sequence several times, compiling into one disconnected Selection
-holding several independent anchor pairs:
-
-```svml
-@beat one @/beat ... @beat three @/beat
-```
-
-Different ids may cross freely; XML-style nesting is not required:
-
-```svml
-@a one @b two @/a three @/b
-```
-
-The specification requires the parser to manage open state per id rather than with a single
-global stack. The following MUST fail:
-
-- an unclosed marker, an orphan close, or mismatched open/close ids;
-- re-opening an id that has not yet been closed;
-- a Selection marker cutting into a v1 speech token;
-- relying on a consumer default span to complete a missing endpoint.
-
-Multiple occurrences of the same id keep source order and stable occurrence identity. Different
-SelectionSets, and different occurrences of the same set, may overlap in physical time.
-
-### 4.3 Optional temporal boundaries
-
-After Slot binding, Dual Text speech projection and NFC normalization, v1 uses the deterministic,
-normatively defined `speech-tokenizer-v1`:
-
-- East Asian ideographs and kana form single-character tokens;
-- other supported script/digit characters form maximal contiguous runs;
-- `'` or `’` is kept only when it sits between two run characters;
-- whitespace, punctuation, emoji and other separator characters produce no lexical token.
-
-Punctuation therefore cannot be timed on its own, and a Selection/Moment cannot cut into a
-single token. The v1 compiler tokenizes after removing the zero-width markers, so if grapheme,
-syllable or phoneme boundaries are ever needed, the same Selection/Moment types can be reused in
-a new timing profile / Script Surface version; v1 should not introduce new body symbols for a
-precision that does not yet exist.
-
-If a Script has `N` Segments, the `k`-th Segment has `mₖ` speech tokens, and `M = Σmₖ`, the
-Semantic Anchor Index has exactly:
-
-```text
-Σ(2mₖ + 2) = 2M + 2N
-```
-
-stable identities. The local order inside each Segment is:
-
-```text
-segment[k].start
-token[k,1].start
-token[k,1].end
-...
-token[k,mₖ].start
-token[k,mₖ].end
-segment[k].end
-```
-
-The Program start and end belong to the ProgramBasis and do not additionally enter the Semantic
-Anchor Index. An empty Segment still has an independent start/end even when the two ultimately
-coincide; any two distinct identities MUST NOT be merged even if they end up on the same frame.
-
-The Locator MUST commit a total mapping covering all `2M + 2N` identities:
-
-```text
-SemanticAnchorIdentity → ProgramPoint(ProgramSpace)
-```
-
-A marker's left/right affinity is resolved to exactly one anchor at Script parse time. Token cut
-points and Segment cut points are entirely equal here: a left-affine marker at the head of a
-Segment takes that Segment's own start cut point and never reaches back to the word end of the
-previous Segment. Downstream receives anchor identities and no longer performs index arithmetic
-of its own.
-
-Location is total: every token of every Segment carries a window, whether it was measured,
-inferred from adjacent characters, or interpolated because the transcript never touched it.
-Reversed windows, windows outside their owning Segment, and windows overlapping their neighbours
-are all reported exactly as measured; that is a fact about the recording, and interpreting it
-belongs to whoever projects it onto a timeline. Location fails only when the Script, the audio
-and the transcript are not the same set of three things, never because of a single timestamp.
-
-Every point MUST exist. A consumer MUST NOT invent points, move points, or borrow points from an
-adjacent occurrence. Inside each Segment the order MUST be non-decreasing:
-
-```text
-segment.start ≤ token₁.start ≤ token₁.end ≤ ... ≤ segment.end
-```
-
-Segments `A` and `B` that are adjacent in source MUST additionally preserve:
-
-```text
-A.start ≤ B.start
-A.end ≤ B.end
-```
-
-This permits hard cuts, overlaps and gaps, while forbidding a Locator from silently ordering an
-entire later-written Segment before an earlier-written one:
-
-```text
-hard cut A.end == B.start
-overlap B.start < A.end
-gap B.start > A.end
-```
-
-If genuine reordering or parallel speech is ever needed, an explicit non-linear narrative model
-should be published; v1 does not let an ordinary Range reverse or vanish under a different
-Locator.
-
-The same Selection yields preview or final-cut intervals on the `CompleteSemanticMap` produced by
-different fulfillments. The Map uses the same identities. Script itself contains no seconds, frame
-numbers or sample points. Script Surface is not bound to a frame rate, sample rate or renderer.
-Once a backend has chosen a physical clock it MUST quantize exactly once and let every consumer
-reuse the same integer boundaries; a change of backend clock does not change this language surface.
-
-Alignment, structural endpoints, projection or frame quantization may make some occurrence end up
-as a zero-length or reversed interval. An implementation MUST produce an explicit diagnostic and
-MUST NOT silently drop it, move Selection endpoints, borrow an adjacent occurrence, or complete
-the range on the author's behalf.
-
-## 5. Moment
-
-A Moment is a complete point declaration, not a Selection with its close omitted:
-
-```svml
-@id!
-~@id!
-```
-
-`!` explicitly marks the type of this name as Moment. It neither needs nor allows a close:
-
-| Form | Affinity | Typical meaning |
-|---|---|---|
-| `@id!` | Right | Word start on the right or structural endpoint on the right |
-| `~@id!` | Left | Word end on the left or structural endpoint on the left |
-
-The default right affinity covers the frequent "starts together with some word" case:
-
-```svml
-I just @pop! laughed my ass out.
-```
-
-`pop` resolves to the onset of `laughed`. Explicit left affinity is written as:
-
-```svml
-I laughed ~@pop! and left.
-```
-
-`pop` resolves to the release of `laughed`. When it sits between two Segments, the left candidate
-is the previous Segment's `end` and the right candidate is the following Segment's `start`; the
-two may be the same point, or different points because of overlap or gap.
-
-Normative rules:
-
-- `@id!` / `~@id!` with the same id may repeat and compile in source order into one Moment; the
- public value of each occurrence is a single resolved `anchorId`.
-- Selection and Moment share the temporal name namespace. One id MUST NOT declare both types;
- `@x! ... @/x` MUST fail with a type conflict.
-- A lone `@x` is always an unclosed Selection open and MUST be an error; the parser MUST NOT
- guess it to be a Moment just because it found no `@/x`.
-- `@id!~` is a redundant and illegal right-affine form; right affinity is written only in the
- canonical form `@id!`.
-- The syntax accepts only ASCII `!`. An unescaped `@id!` (fullwidth exclamation mark) is
- malformed temporal syntax and MUST NOT degrade to body text.
-- Moment and Selection markers may appear in the same positions, including between Segments and
- on the speech side of Dual Text, but neither may cut into a v1 speech token.
-- A Moment only selects one candidate point already present in the Semantic Anchor Index; it adds
- no anchor identity, so the exact count `2M + 2N` does not change with the number of Moments.
-- Moment and Selection reuse the same CompleteSemanticMap and the same one-time frame
- quantization; the only difference is that the final carrier is `Point[]` rather than `Range[]`.
-
-The consumer side keeps two types that never mix:
-
-```text
-NarrativeSelection = { startAnchorId, endAnchorId }[]
-NarrativeMoment = { anchorId }[]
-```
-
-The recommended explicit usage relationships are:
-
-```svml
-during="phrase" <!-- accepts a SelectionSet only -->
-at="pop" <!-- accepts a MomentSet only -->
-at="phrase.start" <!-- start-point projection of each range of a SelectionSet -->
-at="phrase.end" <!-- end-point projection of each range of a SelectionSet -->
-```
-
-`at="phrase"` MUST NOT default to guessing start, and `during="pop"` MUST also be a type error. A
-consumer that needs both a window and a trigger point MUST use two fields/ports rather than build
-a polymorphic `Range | Point | mixed[]` TemporalRef. Whether a port needs `one` or `each` belongs
-to the consumer's cardinality contract and does not change Script's single/multiple occurrence
-syntax.
-
-## 6. Slot
-
-Current implementation status: `parseScript` already implements the parse-first, literal-only
-binding contract below, but the official `<script>` Author Surface does not yet expose a bindings
-input. An unbound Slot currently fails in `main.svml` with `SCRIPT_SLOT_UNBOUND`; public usage
-examples MUST NOT claim that Slot is directly usable before the Surface completes the binding edge.
-
-A Slot is a first-class AST atom recognized by the parser:
-
-```svml
-${product}
-<${product} | ${product_pronunciation}>
-```
-
-A Slot id after NFC MUST consist of 1–64 characters: the first character is a Unicode Letter or
-`_`, and the remaining characters may be Unicode Letter, Mark, Decimal Number, `_` or `-`. A bound
-value is single-line Unicode plain text and MUST NOT contain CR/LF or control characters.
-
-Key safety rules:
-
-- Parse the Script first, then bind Slots; string substitution before parsing is forbidden.
-- A bound value always enters an already existing atom as literal text and MUST NOT inject a
- Segment, Role Cue, Dual Text, Selection, Moment, Slot, comment or escape sequence.
-- An unbound Slot may remain in the structured editing IR, but anything that produces the three
- texts, computes tokens or compiles a timeline MUST fail closed.
-- An SVML v1 writer emits only `${variable}`; an unescaped `[variable]` is just ordinary body text.
-
-## 7. Lexical reservation, escaping and canonical formatting
-
-The source uniformly converts CRLF/CR to LF and applies Unicode NFC. Unescaped `<`, `@` and `${`
-in the source language are reserved syntax starts. Malformed reserved syntax MUST be an error and
-MUST NOT degrade to spoken text.
-
-Body escapes:
-
-```svml
-\@openai
-\<not-a-role>
-\${literal}
-\\
-```
-
-Inside Dual Text there are additionally:
-
-```svml
-\|
-\>
-```
-
-An unknown backslash escape MUST be an error. Parse priority is fixed as:
-
-1. comments, temporal markers and named Segment opens in Script body state;
-2. the current Segment's close in Segment body state;
-3. an inline `<display | speech>` containing one unescaped separating `|` in a Segment body;
-4. a bare `<Role>` in a Segment body satisfying the Role label contract;
-5. any other angle-bracket construct is an error.
-
-Segment names and temporal names use independent namespaces; Selection and Moment share the
-temporal name namespace. All three kinds of id satisfy:
-
-```text
-[a-z][a-z0-9_-]{0,63}
-```
-
-Source indentation, newlines and runs of whitespace are all authoring layout: they take no part
-in recognizing a Segment, a Role Cue or a turn, and they are not caption line-break or pause
-instructions. The compiled projections preserve the necessary separation at atom boundaries and
-normalize layout whitespace; caption cue segmentation and hard line breaks are decided by the
-outer Caption Program.
-
-Structural tags and comments do not depend on occupying a line of their own to be parsed. The
-canonical formatter puts each structural tag and each comment on its own line, indents Segments
-by two spaces and spoken content inside a Segment by four spaces, and leaves one blank line
-between Segments. The Narrative IR obtained by re-parsing before and after formatting MUST have
-identical semantic content; the only things allowed to change are the source range and the source
-hash.
-
-Comments use `<!-- ... -->`. They may span lines but MUST NOT nest, and MUST NOT be written inside
-Dual Text, a Slot, a temporal marker, a Role Cue or a structural tag. A comment may appear wherever
-layout whitespace is allowed between atoms and is treated as layout whitespace; the canonical
-formatter gives it its own line. A comment enters the semantic content of no projection, token,
-Selection, Moment or hash; a separate source hash may be computed when a reproducible source
-identity is needed.
-
-## 8. Narrative IR and the consumer contract
-
-The parse result preserves at least:
-
-- Segment order, ids and independent start/end identities;
-- each Segment's ordered spoken turns, together with each turn's stable identity, optional Role
- Cue, token range and source range;
-- plain / Dual Text / Slot atoms and their source ranges;
-- Caption display Atoms, their internal ordered display Words, and the independent explicit
- mapping from an Atom to one or more speech tokens;
-- one or more ordered occurrences per Selection, and each occurrence's resolved
- `startAnchorId` / `endAnchorId`;
-- one or more ordered occurrences per Moment, and each occurrence's resolved `anchorId`;
-- the Parser/Source Map separately preserves affinity and token/source ranges; public values do
- not carry them;
-- the complete ordered `CaptionDisplaySequence`, an independent `CaptionCorrespondence`, and a
- `CaptionDisplayWordSubset` for every explicit Selection. Captions need not, and MUST NOT,
- reverse-engineer word indices from public Selections.
-
-An outer Program's `role="label"` selector may project its own word subset out of the full word
-set above; several Turns with the same label are several ordered, mutually non-adjacent hits. It
-is an explicit query internal to the Caption author surface, not an implicit caption behaviour of
-Role Cue, and it establishes no speaker entity.
-
-SelectionSet/MomentSet are consumer boundaries, not sub-objects of a Segment. The typical external
-relationships are:
-
-```text
-Script ──compile──> Narrative IR ──project──> dialogue / speech
- ├──resolve──> NarrativeSelection[] / NarrativeMoment[]
- └──project──> CaptionDisplaySequence
- + CaptionCorrespondence
- + CaptionDisplayWordSubset[]
-
-NarrativeSelection A ──during──> consumer.port_1
-NarrativeMoment P ──at──────> consumer.port_2
-CaptionDisplayWordSubset C ──words──> caption.Program
-```
-
-One Ranking, B-roll or other node may have several dynamic ports, each accepting different
-Selections/Moments; the same set may also be reused by several consumers. A future Graph syntax
-may write these typed edges as `during=` / `at=`, but they MUST reference resolved names or
-endpoint projections and MUST NOT accept vague natural-language locators.
-
-Caption multi-style, region, cue segmentation, annotation, mute and layout all belong to the
-Caption Program. If a Caption consumer requires mutually exclusive token ownership, an overlap
-MUST be a compile error for that consumer; this is not Script forbidding Selections to overlap.
-If Caption ever supports overlay, only the Caption Program is extended, not the body syntax.
-
-The current Caption Program uses one default Style covering every visible word, then applies whole
-Style replacements from Roles or explicit Selections in source order, with the last matching rule
-winning. A Turn without a Role automatically keeps the default Style; modifying one local interval
-does not require the author to write its complement. The complete rules are in
-[`caption-program.md`](./caption-program.md).
-
-## 9. What v1 explicitly does not express in Script
-
-The following capabilities are not a missing seventh inline marker:
-
-| Requirement | v1 owner |
-|---|---|
-| Absolute seconds, frame numbers, sample points, arbitrary sub-frame positions | Runtime/Program outside Script |
-| Phoneme- or grapheme-level cut points inside a token, punctuation-specific timing | A later timing profile or a new version |
-| Preview duration of an empty Segment | Media, a generator or an explicit Program |
-| Simultaneous overlapping speech | Multi-track media/generation Program; the Base Script remains a sequential semantic axis |
-| Laughter, coughing, ambience, SFX, beeps | Audio/generation/SFX Program |
-| Pure display text with no speech timing basis | Text, Deck or Caption Program |
-| Caption styling, manual line breaks, motion, on-screen action | The corresponding Program referencing a Selection |
-| Tone, emotion, timbre, stress, rate, SSML | Speech/Generation Program |
-| Consumers, nodes, ports, DAG wiring | SVML regions outside `<script>` |
-
-These boundaries keep Script a "script you can read directly", while not preventing full SVML from
-ultimately compiling a deterministic DAG and timeline.
-
-## 10. Completeness and the extension test
-
-Classify a new requirement in the following order:
-
-1. The display text differs from the actual pronunciation: Dual Text.
-2. The speaker prefix of the dialogue export differs: Role Cue.
-3. The semantic time range differs: Selection.
-4. The semantic instant differs: Moment.
-5. The sequential media/spoken block differs: Segment.
-6. The runtime text input differs: Slot.
-7. The caption, picture, audio, generation or consumption behaviour differs: a Program outside
- Script.
-
-As long as a requirement falls into one of these seven, no new inline delimiter may be invented.
-If none of the seven can express it losslessly, first prove that it is script semantics rather
-than a Program, then handle it as a new version proposal; an existing parser MUST NOT be made to
-"support it along the way".
-
-## 11. Implementation acceptance
-
-An implementation may declare Script Surface v1 support only when it satisfies all of the
-following evidence at once:
-
-1. every positive example, negative example, projection and source-map golden fixture in this
- document;
-2. parser, formatter, Narrative IR and the three projections depend on no provider and no
- renderer;
-3. SelectionSet supports closure, crossing, disconnected occurrences and left/right affinity;
-4. MomentSet supports `@id!` / `~@id!`, single/multiple occurrences and the `at` type gate;
-5. Dual Text keeps one auditable source map across speech, caption and temporal mapping;
-6. Slot uses parse-first, literal-only binding and cannot inject syntax;
-7. unknown versions, unknown reserved syntax, mixed temporal types and partial dual atoms all
- fail closed.
-8. the canonical serialization, digest, complete Map and adjacent-Segment double monotonicity
- constraint for the `2M + 2N` identities all have golden fixtures; hard cut, overlap, gap,
- empty Segment and coincident endpoints are all covered.
diff --git a/docs/guide/spec/spatial-layout.md b/docs/guide/spec/spatial-layout.md
deleted file mode 100644
index 594ac578..00000000
--- a/docs/guide/spec/spatial-layout.md
+++ /dev/null
@@ -1,472 +0,0 @@
----
-title: Shared Spatial Layout
-description: executable foundational authority for shared Spatial geometry. `@narratage/spatial`.
----
-
-# Shared Spatial Layout
-
-Status: executable foundational authority for shared Spatial geometry. `@narratage/spatial`
-implements CanvasSpace, Frame/Point/Path, IntrinsicExtent, ContentFit and pure fitting; Film now
-consumes the same explicit CanvasSpace as Track layout. Text, Media and all four Ranking witnesses
-execute, and the repository-internal Visual IR compatibility audit passes. Public package ABI
-publication remains separate.
-
-## 1. Conclusion
-
-The shared spatial system owns coordinate geometry, not Text, Media or CSS. Its stable vocabulary is:
-
-```text
-CanvasSpace exact output coordinate system
-SpatialPoint one point in that system
-SpatialFrame one positive rectangular region
-SpatialPath one explicit vector path
-IntrinsicExtent one content item's natural two-dimensional size/aspect
-ContentFit how an extent is sized and aligned against a Frame
-FittedContent the deterministic result of that fit
-```
-
-When a visual component consumes a `SpatialFrame`, that frame plays the role of its
-**Placement Frame**: where the component is placed and where frame-level Paint may exist. Every
-actual text/media sample then derives its own **Content Frame** from its intrinsic dimensions and a
-`ContentFit`.
-
-These are the two frames. They are not always nested:
-
-```text
-contain cover
-
-Placement Frame Content Frame
-┌──────────────┐ ┌──────────────────┐
-│ ┌────────┐ │ │ ┌────────────┐ │
-│ │Content │ │ │ │ Placement │ │
-│ │ Frame │ │ │ │ Frame │ │
-│ └────────┘ │ │ └────────────┘ │
-└──────────────┘ └──────────────────┘
-```
-
-Calling the second one an "inner box" is therefore incorrect. Under `cover`, zoom or local motion,
-the Content Frame may exceed the Placement Frame.
-
-One Placement Frame may contain several explicit samples. A solid backing has no Content Frame. A
-blurred sample and a foreground sample each have their own Content Frame, even when they reference
-the same Artifact. There is no single global "inner frame" shared by every layer.
-
-## 2. Boundary
-
-The focused video-domain package is `@narratage/spatial`. It owns pure types,
-validators, author Surfaces for reusable geometry and deterministic geometry Producers. It does not
-own:
-
-- temporal windows or intrinsic media playback;
-- text wrapping or glyph measurement;
-- media decoding, stream selection or container rotation;
-- backgrounds, blur, opacity, color, borders or shadows;
-- z-order, transitions or local motion;
-- Composition, Runtime, Provider or Core behavior.
-
-Those facts meet Spatial only through graph edges. For example, Media Inspection selects one visual
-stream and produces its display-oriented extent; a Media package explicitly converts that extent
-into `IntrinsicExtent` and gives it to Spatial fitting. Spatial never opens an MP4 or guesses which
-stream is primary.
-
-Installing or revising this package must not change Core. The Core sees ordinary typed Records and
-Operations.
-
-## 3. CanvasSpace
-
-`CanvasSpace` establishes the exact coordinate system before any Track is authored:
-
-```ts
-type CanvasSpace = {
- readonly contract: "svml.canvas-space@1";
- readonly widthPx: number;
- readonly heightPx: number;
- readonly origin: "top-left";
- readonly xDirection: "right";
- readonly yDirection: "down";
- readonly pixelAspect: "square";
-};
-```
-
-Width and height are positive integers. Color, clear Paint, frame rate and duration are not spatial
-facts and remain outside this value.
-
-The Canvas must be declared before Tracks and later connected to Film/Composition as the same graph
-value. The current Film Surface hides width and height inside the final Film declaration, which is
-too late for exact Text and Media layout. Its rewrite must separate Canvas declaration from final
-Track assembly rather than copying width/height into every Track Recipe.
-
-An intended source shape is:
-
-```svml
-<space:Canvas id="vertical" width="1080" height="1920"/>
-
-<!-- Track declarations consume vertical or Frames derived from it. -->
-
-<film:Film id="main" canvas={vertical} space={program} appearance={studio.film.main}>
- ...
-</film:Film>
-```
-
-The exact Surface spelling remains pre-freeze. The one-value dataflow is normative.
-
-## 4. SpatialFrame construction
-
-A resolved `SpatialFrame` is a finite rectangle in Canvas pixel coordinates:
-
-```ts
-type SpatialFrame = {
- readonly contract: "svml.spatial-frame@1";
- readonly xPx: number;
- readonly yPx: number;
- readonly widthPx: number;
- readonly heightPx: number;
-};
-```
-
-Width and height must be positive. `xPx` and `yPx` may be negative or extend beyond Canvas. An
-off-canvas Frame is useful for entry/exit motion and must not be silently clamped, intersected or
-rejected. Canvas rasterization determines which pixels are visible. Clamping is always an explicit
-projection choice.
-
-Author inputs may use Canvas-relative percentages or pixels. They compile into this one resolved
-form before a consumer lays out its content. Percentages are relative to the explicitly connected
-parent Frame's width/height, not to whichever DOM container a renderer happens to create.
-
-Three author components cover the normal construction meanings without a mode field:
-
-1. `Frame`: four explicit edges inside a parent Canvas/Frame;
-2. `AnchoredFrame`: one target point, one width/height and one of nine self anchors;
-3. `AspectFrame`: one target point, one primary size and one explicit or connected aspect ratio.
-
-Conceptually:
-
-```svml
-<space:Frame id="safe" within={vertical}
- left="6%" top="4%" right="94%" bottom="96%"/>
-
-<space:AnchoredFrame id="card" within={safe}
- x="50%" y="78%" width="82%" height="28%" anchor="center"/>
-
-<space:AspectFrame id="sticker" within={safe}
- x="100%" y="100%" width="32%" aspect={portrait.extent} anchor="bottom-right"/>
-```
-
-`Frame` does not mean a visible box. It produces geometry only. `AspectFrame` may consume a generic
-`IntrinsicExtent` edge; it does not depend on Media.
-
-The nine canonical anchor names are:
-
-```text
-top-left top-center top-right
-middle-left center middle-right
-bottom-left bottom-center bottom-right
-```
-
-The compiled math is point attachment, not nine hard-coded layout algorithms:
-
-```text
-frame.x = target.x - frame.width × selfAnchor.x + offset.x
-frame.y = target.y - frame.height × selfAnchor.y + offset.y
-```
-
-Each named anchor is merely a readable `(0 | 0.5 | 1, 0 | 0.5 | 1)` pair. An advanced Surface may
-expose arbitrary normalized anchor points without changing the compiled contract.
-
-Safe area, left column, title region and product region are ordinary named Frames derived from a
-Canvas or parent Frame. They are not reserved global tokens.
-
-## 5. IntrinsicExtent
-
-`IntrinsicExtent` is the minimal generic fact needed to preserve content shape:
-
-```ts
-type IntrinsicExtent = {
- readonly contract: "svml.intrinsic-extent@1";
- readonly widthPx: number;
- readonly heightPx: number;
-};
-```
-
-The values are the content's positive display-oriented natural pixel extent. Their ratio is all that
-`contain`, `cover`, `fit-width`, `fit-height` and `stretch` need; `native` and `scale-down` also use
-their absolute relation to Canvas pixels. Media may derive them after applying container rotation
-and sample-aspect information. Text may derive them from exact font shaping. A vector component
-must choose an explicit nominal raster extent rather than pretending arbitrary view-box units are
-Canvas pixels. None stores the source Artifact, font, component family or upstream digest inside
-`IntrinsicExtent`; those relationships remain visible on graph edges and Derivations.
-
-## 6. ContentFit and the two-frame equations
-
-`ContentFit` chooses sizing, alignment and an optional alignment constraint:
-
-```ts
-type ContentFit = {
- readonly contract: "svml.content-fit@1";
- readonly sizing:
- | "contain"
- | "cover"
- | "fit-width"
- | "fit-height"
- | "native"
- | "scale-down"
- | "stretch";
- readonly framePoint: { readonly x: number; readonly y: number };
- readonly contentPoint: { readonly x: number; readonly y: number };
- readonly offsetPx: { readonly x: number; readonly y: number };
- readonly constraint: "bounded" | "free";
-};
-```
-
-Points are normalized in their own Frames. `framePoint` says where in the Placement Frame the
-chosen content point wants to land. `contentPoint` says which intrinsic source point is important.
-This is more expressive than the old single `focalX/focalY`, which silently used the same coordinate
-on both sides. Both point coordinates must lie in `[0,1]`; deliberate displacement beyond those
-bounds uses `offsetPx` and, when it must remain out of bounds, `constraint="free"`.
-
-Given Placement Frame `(Vx, Vy, Vw, Vh)` and intrinsic extent `(Sw, Sh)`, the aspect-preserving
-base scales are:
-
-```text
-contain min(Vw / Sw, Vh / Sh)
-cover max(Vw / Sw, Vh / Sh)
-fit-width Vw / Sw
-fit-height Vh / Sh
-native 1
-scale-down min(1, contain-scale)
-```
-
-For these modes:
-
-```text
-Cw = Sw × scale
-Ch = Sh × scale
-```
-
-`stretch` deliberately abandons source aspect and produces `(Cw, Ch) = (Vw, Vh)`. It is explicit
-because mature editors permit distortion; no other sizing mode may distort.
-
-Before constraint:
-
-```text
-Cx = Vx + Vw × framePoint.x - Cw × contentPoint.x + offset.x
-Cy = Vy + Vh × framePoint.y - Ch × contentPoint.y + offset.y
-```
-
-For `free`, these coordinates are final. For `bounded`, each axis is minimally clamped by one law:
-
-- when content is smaller, keep the complete Content Frame inside the Placement Frame;
-- when content is larger, keep the complete Placement Frame covered by the Content Frame.
-
-For the horizontal axis this means:
-
-```text
-Cw <= Vw: clamp Cx to [Vx, Vx + Vw - Cw]
-Cw >= Vw: clamp Cx to [Vx + Vw - Cw, Vx ]
-```
-
-The vertical law is identical. This supports true focal placement such as “source point `(0.8,
-0.35)` should appear at the center of the Frame” without permitting an accidental uncovered strip.
-
-The usual nine alignments are shorthand where `framePoint === contentPoint`. Centered contain and
-centered cover are both `(0.5, 0.5)` on each side. `top-left` is `(0,0)` on each side. An author who
-needs the content deliberately outside or the Placement Frame deliberately uncovered chooses
-`free`; the Runtime never makes that creative decision.
-
-`FittedContent` contains the derived Content Frame and no duplicated upstream lineage:
-
-```ts
-type FittedContent = {
- readonly contract: "svml.fitted-content@1";
- readonly contentFrame: SpatialFrame;
-};
-```
-
-When clipping to the Placement Frame, the corresponding source sampling window is mechanically
-derived from the intersection of Placement and Content Frames. It is not a third authored box. A
-post-fit rotation or perspective transform may make the sampling boundary non-rectangular and is
-therefore applied after this base layout.
-
-## 7. Fit is not motion and not occupancy
-
-A fit calculation produces the base Content Frame. Zoom, pan, rotation, perspective and entrance
-motion transform that result later. They do not mutate the `ContentFit` or source aspect.
-
-This prevents the old ambiguity where `zoom` was embedded inside `contain`/`cover` and could quietly
-invalidate the claimed fit. If a Ken Burns motion zooms a contained image until it crops, the base
-fit remains `contain` and the later transform explicitly creates the crop.
-
-Spatial sizing is also unrelated to intrinsic temporal occupancy. A video may be `contain` in space
-while it plays once, loops or stretches in time. The same image has no intrinsic temporal playback
-mode merely because it has an intrinsic spatial extent.
-
-## 8. Clipping and frame shape
-
-Spatial geometry never decides whether pixels outside a Placement Frame are visible. The consuming
-component owns an explicit clip policy:
-
-```text
-none owned content may paint outside the Placement Frame
-placement-frame rectangular or rounded Frame clip
-owned-shape explicit shape/path owned by the same component
-```
-
-Corner radius, ellipse/path masks, border and shadow are presentation, not coordinates. A Text
-glow may legitimately overflow its Placement Frame; a Media card normally clips every local sample
-to the same rounded Placement Frame. No default is inferred from content type by Runtime.
-
-An owned shape may clip only elements in the same component contribution. It cannot mask or sample
-a sibling Track.
-
-## 9. Transparent, solid, blurred and alternate backings
-
-The area of a Placement Frame not painted by a Content Frame is not itself spatial meaning.
-Media/Text presentation supplies an ordered list of explicit owned layers.
-
-### Transparent
-
-No backing layer is emitted. Transparent pixels naturally reveal lower Tracks during Composition.
-This is not cross-Track sampling: the Track contributes no pixel there.
-
-### Solid or gradient
-
-A frame Paint layer fills the Placement Frame. It has no source input, IntrinsicExtent or Content
-Frame.
-
-### Blurred copy of the same source
-
-The source is connected explicitly a second time as another local sample:
-
-```text
-same source Artifact ─┬─> cover ContentFit -> blur/tone -> backing layer
- └─> contain ContentFit -> foreground layer
-```
-
-Both samples share one Placement Frame but have independent Content Frames and effects. Artifact
-collection deduplicates identical bytes; rendering still performs two explicit samples. The blur
-implementation must edge-extend/overscan before clipping so the Gaussian kernel cannot create an
-unintended transparent or dark border.
-
-### Another image or video
-
-The alternate source enters on its own graph edge and receives its own Inspection, IntrinsicExtent,
-ContentFit and, for timed media, occupancy. It is not a string path hidden in a Style Recipe.
-
-The general Media author model is therefore ordered local layers rather than special
-`foreground/background` fields:
-
-```svml
-<media:Item id="portrait-card" frame={card} clip={rounded-card}>
- <media:Paint recipe={studio.media.card-base}/>
- <media:Layer source={portrait} fit={cover} effects={blurred}/>
- <media:Layer source={portrait} fit={contain}/>
-</media:Item>
-```
-
-Changing the second source to another image expresses an alternate backing. Removing the first two
-children expresses transparent backing. Each layer may lower to its own absolute-stack Present when
-peer Track interleaving is required; an authoring Track never becomes an implicit stacking context.
-
-The complete Item/Sequence author model is specified in [`media-track.md`](./media-track.md). The
-independent depth-stack collection model is specified in [`deck-track.md`](./deck-track.md). Both
-reuse the explicit layer and edge laws without making Deck a Media mode.
-
-### What remains forbidden
-
-`backdrop-filter` over the already accumulated Composition is not one of these cases. It would make
-the result depend on arbitrary lower Tracks. To blur another visual, the component must receive
-that visual/media explicitly. To blur a completed lower composite, a higher-order component must
-explicitly own and materialize that composite first.
-
-## 10. Text, Ranking and Caption use
-
-Spatial provides common geometry without forcing one layout model:
-
-- Point Text consumes a `SpatialPoint` and derives its Content Frame from exact text extent;
-- Area Text consumes a Placement Frame, then performs package-owned paragraph flow inside it;
-- Path Text consumes an explicit `SpatialPath`;
-- Media consumes a Placement Frame and derives one Content Frame per sample;
-- Ranking consumes one Placement Frame and owns board, slot and staging-area layout inside it;
-- Comment Sticker consumes a Placement Frame and owns its avatar/text/internal layout;
-- Caption may consume a Placement Frame, but Cue/Atom wrapping remains Caption-owned.
-
-Text's layout box and Media's Content Frame are not the same semantic Type merely because both are
-rectangles. They may use the same pure geometry helpers while remaining package-owned resolved
-facts.
-
-## 11. Graph dataflow and no metadata propagation
-
-```text
-CanvasSpace + Frame Projection
- │
- ▼
- SpatialFrame ───────────────────────────────┐
- │
-Media Inspection -> display extent -> IntrinsicExtent│
- ├─> fit -> FittedContent
-authored ContentFit ─────────────────────────────────┘ │
- ▼
- package Paint/motion/clip -> VisualTrack
-```
-
-Every dependency is a graph edge. `BlobArtifact` does not gain placement fields. `SpatialFrame`
-does not gain source, role, timing or Artifact digests. `VisualTrack` does not gain `focal`, B-roll
-or Text fields. Affinity is validated by the multi-input fitting/lowering Operation rather than by
-copying lineage metadata through every value.
-
-## 12. Legacy findings retained and retired
-
-Retained:
-
-- normalized author placement and source-aspect-aware anchored placement;
-- independent source fitting and temporal occupancy;
-- contain, cover, fit-width and fit-height equations;
-- independent foreground and backing sampling;
-- source focal alignment and explicit attachment points;
-- optional transparent, solid and self-blurred backings;
-- local motion over sampling/placement.
-
-Retired:
-
-- one locator union mixing regions, VLM inference, avoidance, path signals and fixed boxes;
-- nine-grid codes as separate algorithms;
-- one `focalX/focalY` reused as both source and target point;
-- a media box style treating exactly one foreground and one special backing as the universal model;
-- embedding zoom into fit;
-- renderer-specific CSS and FFmpeg formulas as two sources of truth;
-- hidden `object-fit`, `object-position` or `backdrop-filter` defaults.
-
-## 13. Implementation and acceptance order
-
-1. **Implemented:** add the focused `@narratage/spatial` package with `CanvasSpace`, Frame/Point/Path,
- `IntrinsicExtent`, `ContentFit` and pure fit validation;
-2. **Implemented:** split Canvas declaration from Film assembly and connect the same Canvas value to Track layout and
- Composition;
-3. **Implemented:** migrate one simple Text Area and one simple Media still image through shared geometry;
-4. **Implemented:** add exact browser tests for every fit and alignment law;
-5. **Implemented:** prove ordered transparent/color/self-blur/alternate-source Media layers;
-6. **Implemented:** migrate all four Ranking components through explicit shared Frames; Comment
- Sticker remains deferred until it has its own design;
-7. only then freeze the Spatial Types and related Visual IR behavior.
-
-The test matrix must include:
-
-- portrait, landscape and square intrinsic extents against portrait, landscape and square Frames;
-- every sizing mode, including explicit distortion and scale-down;
-- all nine equal-point alignments plus unequal source/frame focal points;
-- bounded versus free alignment;
-- negative/partially/fully off-canvas Placement and Content Frames;
-- parent Frame percentages and pixel offsets;
-- source rotation/sample-aspect normalized before Spatial;
-- transparent, solid, same-source blur and alternate image/video layers;
-- independent backing and foreground Content Frames;
-- rounding/raster behavior locked by the selected renderer implementation;
-- no Core, Runtime, Film-family, Text-family or Media-family discriminator in the shared Types.
-
-The pure geometry matrix and real Chromium witness now cover every sizing mode, all nine equal-point
-alignments, unequal bounded/free focal points and partially/fully off-Canvas Frames. Media
-normalization additionally materializes container rotation and non-square sample aspect into a
-square-pixel visual before deriving `IntrinsicExtent`. Ranking now consumes the same explicit Frame
-geometry without adding a family discriminator. The final Visual IR compatibility audit now passes;
-Spatial's own author Surface remains pre-release, and no package-local percentage vocabulary should
-grow in the meantime.
diff --git a/docs/guide/spec/track-authoring.md b/docs/guide/spec/track-authoring.md
deleted file mode 100644
index b97c3b26..00000000
--- a/docs/guide/spec/track-authoring.md
+++ /dev/null
@@ -1,476 +0,0 @@
----
-title: SVML Video Track Authoring Model
-description: executable temporal authority shared by the official Track packages. The focused.
----
-
-# SVML Video Track Authoring Model
-
-Status: executable temporal authority shared by the official Track packages. The focused
-`@narratage/temporal` package implements strict occurrence expansion, exact rational window
-projection and triggered sibling schedules. Shared Spatial, Text, Ranking, Media,
-Audio and Screen Overlay are now specified in [`spatial-layout.md`](./spatial-layout.md),
-[`typography-track.md`](./typography-track.md), [`ranking-track.md`](./ranking-track.md),
-[`media-track.md`](./media-track.md), [`deck-track.md`](./deck-track.md),
-[`audio-track.md`](./audio-track.md) and
-[`screen-overlay.md`](./screen-overlay.md). The independently designed
-`@narratage/comment-sticker` package now uses the same Temporal and Spatial foundations and lowers
-to one peer VisualTrack. This is deliberately not a compatibility promise for a released public
-ABI.
-
-## Purpose
-
-Official video packages need one shared discipline without one universal Track Program. Text,
-Media, Ranking and Comment Sticker own different author meanings, but they must not independently
-invent incompatible ways to answer the same temporal questions.
-
-This document governs package Programs and their deterministic lowering. It does not add video
-meaning to Core, Runtime, Composition or a Provider. The public terminal waist remains
-[`VisualTrack` and `AudioTrack`](./track-composition.md).
-
-The complete package-authoring separation is:
-
-```text
-content
-temporal source
-occurrence expansion
-window projection
-optional sibling sequencing
-intrinsic-material occupancy
-spatial source and layout
-presentation
-```
-
-These are independent authoring concerns, not mandatory fields in every package and not fields on
-the public Track contract.
-
-## 1. Temporal pipeline
-
-```text
-Narrative Selection ──> { start, end } occurrences ─┐
-Narrative Moment ─────> { cue } occurrences ────────┼─> occurrence expansion
-Program singleton ────> { program.start/end } ──────┘
- │
- ▼
- Temporal Window Projection
- │
- ▼
- raw directed candidate windows
- │
- ▼
- ProgramSpace intersection + validation
- │
- ▼
- frame-exact projected windows
- │
- ▼
- optional package-owned sibling sequencing
- │
- ▼
- final allowed windows
- │
- ▼
- intrinsic-material occupancy / playback
- │
- ▼
- render intervals + source-time sampling
-```
-
-The layers answer different questions:
-
-| Layer | Question |
-|---|---|
-| anchor location | At which physical frame is an authored anchor? |
-| occurrence expansion | Does one authored item consume exactly one occurrence or every occurrence? |
-| window projection | Which ProgramSpace interval may this occurrence occupy? |
-| sibling sequencing | How do related items hand off, overlap or close one another? |
-| occupancy | How does a product with intrinsic duration sample itself inside its final window? |
-
-No `full`, `fixed`, `relative`, `from`, `until`, `fill` or `fit` union may collapse these layers
-again. An author Surface may offer readable shorthand, but it must lower to the separated model.
-
-## 2. Temporal sources provide points, not renderable windows
-
-A Selection occurrence contributes an ordered pair of located points:
-
-```ts
-type LocatedSelectionOccurrence = {
- readonly id: string;
- readonly start: FramePoint;
- readonly end: FramePoint;
-};
-```
-
-A Moment occurrence contributes one located point:
-
-```ts
-type LocatedMomentOccurrence = {
- readonly id: string;
- readonly cue: FramePoint;
-};
-```
-
-A Program source is a singleton occurrence. Every projection environment also has the two exact
-ProgramSpace boundaries `program.start` and `program.end`.
-
-Selection location must preserve Script occurrence order and the two authored anchor identities. It
-must not sort, merge, clip, exchange or take the absolute difference of the points. A located
-Selection whose `end` precedes its `start` remains valid point evidence. Only a consumer creates a
-window and therefore only the projection layer can decide whether the points it actually consumes
-form a valid interval.
-
-A component-owned exact time product, such as an already timed Caption Cue, may enter the later
-frame-window stages directly. It does not need to pretend that its own timing came from a Narrative
-Selection.
-
-## 3. Point expressions and Temporal Window Projection
-
-The shared projection algebra is conceptually:
-
-```ts
-type TemporalPointExpression =
- | { readonly ref: "program.start"; readonly offset?: Duration }
- | { readonly ref: "program.end"; readonly offset?: Duration }
- | { readonly ref: "selection.start"; readonly offset?: Duration }
- | { readonly ref: "selection.end"; readonly offset?: Duration }
- | { readonly ref: "moment.cue"; readonly offset?: Duration }
- | { readonly ref: "absolute"; readonly at: Duration };
-
-type TemporalWindowProjection = {
- readonly start: TemporalPointExpression;
- readonly end: TemporalPointExpression;
-};
-```
-
-The serialized contract will use explicit units for durations. Seconds, milliseconds and integral
-frames are distinct author inputs; conversion to a rational ProgramSpace must have one deterministic
-boundary-quantization rule. `base_start` and `base_end` are retired names because the current
-architecture has no privileged Base Track.
-
-Common author meanings are ordinary expressions:
-
-| Meaning | Projection |
-|---|---|
-| complete Program | `[program.start, program.end]` |
-| identity Selection | `[selection.start, selection.end]` |
-| three seconds from Selection start | `[selection.start, selection.start + 3s]` |
-| three seconds after Selection | `[selection.end, selection.end + 3s]` |
-| persist from Selection start | `[selection.start, program.end]` |
-| lead two seconds into Selection start | `[selection.start - 2s, selection.start]` |
-| three seconds after a Moment | `[moment.cue, moment.cue + 3s]` |
-| first three seconds | `[program.start, program.start + 3s]` |
-| absolute advanced window | `[absolute(2.5s), absolute(6s)]` |
-
-`full`, `manual`, `fixed` and `relative` may remain presentation words in an authoring UI, but they
-are not distinct compiled timing laws.
-
-### Semantic binding may deliberately ignore local coordinates
-
-An author may bind an item to a Selection over the opening words while projecting it to
-`[program.start, program.start + 3s]`. The Selection still contributes author meaning, occurrence
-identity, cardinality and dependency truth even though its located start/end coordinates are not
-used by that projection.
-
-This is valid with `one`. It is deliberately restricted with `each`, as defined below, because
-repeating an occurrence-invariant projection would otherwise manufacture identical windows.
-
-## 4. Occurrence expansion is strict and precedes projection
-
-The old shape `{ kind: "single", cardinality: "exactly_one" }` repeats one fact twice. The executable
-contract has two non-overlapping choices:
-
-```ts
-type OccurrenceExpansion =
- | { readonly kind: "one" }
- | { readonly kind: "each" };
-```
-
-`each` means compile-time expansion, not media looping. Author Surfaces expose it through their
-own explicit occurrence setting and lower to this exact value.
-
-### `one`
-
-`one` has the following exact law:
-
-1. locate the complete source occurrence set;
-2. require its cardinality to equal one before evaluating the projection;
-3. reject zero or more than one occurrence;
-4. never choose `first`, `last`, `best`, nearest or earliest;
-5. project that one occurrence exactly once.
-
-A projection that ignores Selection/Moment coordinates is normally paired with `one`. Program-only
-timing uses the intrinsic singleton Program source and therefore also lowers as `one` without an
-author-visible cardinality setting.
-
-### `each`
-
-`each` has the following exact law:
-
-1. require one or more located source occurrences;
-2. preserve source occurrence identity and source order;
-3. instantiate exactly one projected occurrence for every source occurrence;
-4. evaluate the same projection independently in each occurrence's local point environment;
-5. derive stable result identity from the authored item id and occurrence id, never an incidental
- sorted index;
-6. never merge, deduplicate, sort by time, pick a winner or silently discard an occurrence;
-7. fail the whole deterministic Operation when any occurrence cannot produce a valid window.
-
-For more than one source occurrence, at least one projection boundary must reference a local point
-(`selection.start`, `selection.end` or `moment.cue`). A projection using only Program or absolute
-points is occurrence-invariant and is rejected under `each`: it would create several semantically
-different instances with the same physical window and no temporal distinction. The author should
-use `one`, bind the Program singleton, or author separate explicit items.
-
-Projection is one-to-one. One occurrence cannot fan out into several windows. An author who wants
-two windows from one Selection or Moment writes two items or uses a package-owned component that
-explicitly constructs those two items.
-
-Set aggregation is not a third generic expansion mode. A visibility mask or another genuinely
-set-valued component may define a package-owned set consumer, but ordinary visual items must not use
-that exception to regain hidden merge, union or deduplication behavior.
-
-Caption `Mute` is one concrete package-owned example. It unions explicitly referenced
-`CaptionDisplayWordSubset` values inside `CaptionProgram`; it does not alter generic occurrence
-expansion, manufacture projected windows, or enter Core.
-
-### Cardinality does not decide overlap
-
-`each` says only how an occurrence set expands. It does not say whether the resulting windows may
-overlap, whether adjacent windows should join, or whether one item closes at the next item's cue.
-Those questions cannot be answered before projection and may also depend on package semantics.
-
-After projection, a consuming package must choose and validate one explicit relation appropriate to
-its component, for example:
-
-- `independent`: overlapping windows are meaningful and remain independent;
-- `disjoint`: any overlap between occurrences of the same authored item is an error;
-- a package-owned Sequence or Deck policy that derives final sibling windows and handoffs.
-
-This relation is not part of `OccurrenceExpansion` and is not owned by Core. Official simple Text,
-Comment Sticker and similar repeated items should normally require disjoint projected windows.
-Media groups and Ranking components may own stronger, explicit sibling rules. No package receives a
-hidden global auto-stitch behavior.
-
-### Triggered stages are sibling sequencing, not another temporal source
-
-Ranking, Tier, slides and other progressive components often have only an ordered set of trigger
-Moments. Reaching trigger `i` enters state `i`; reaching trigger `i + 1` hands off to the next state.
-This case uses the same point location and window projection. It does not justify a staging Type in
-Core or another Script marker.
-
-`N` trigger points alone cannot bound `N` finite stages: the last stage still needs one terminal
-boundary. A triggered component therefore owns an explicit outer window `[B, O)` and an explicit
-stage-terminal point `T`. `T` may reference `O`, but the package must not silently assume that it
-does. Keeping them separate lets a Ranking component finish its last item at `T` and leave the
-fully-settled board visible until `O`. Given located and frame-quantized trigger points
-`p1 ... pN`, the package may derive:
-
-```text
-stage 1 [p1, p2)
-stage 2 [p2, p3)
-...
-stage N [pN, T)
-settled suffix [T, O) (when T < O)
-```
-
-The interval before `p1` is inactive, not an implicit stage zero. If a design needs an initial
-visible state from the enclosing start to `p1`, that state must be explicit and produces `N + 1`
-stages. A renderer must never infer it from the presence of a board or background.
-
-This pattern has two related but distinct package-owned lowerings:
-
-```text
-outer component [B, O)
-cumulative item 1 [p1, O)
-cumulative item 2 [p2, O)
-cumulative item 3 [p3, O)
-
-exclusive stage 1 [p1, p2)
-exclusive stage 2 [p2, p3)
-exclusive stage 3 [p3, T)
-settled suffix [T, O) when T < O
-```
-
-The cumulative form is ordinary `each` projection of `[moment.cue, outer.end]` with an explicit
-overlap policy. It naturally implements a Tier board: after `p2`, both item 1 and item 2 are active;
-the visible prefix itself is the current logical state. No materialized board snapshot is required.
-
-The exclusive form starts from the same candidates and applies a successor-handoff policy: each
-candidate's final end becomes the next candidate's start, while the last ends at `T`. This can drive
-a Ranking staging area that shows exactly one current item while already-settled items remain as a
-separate cumulative prefix. When `T < O`, the suffix has no current item and shows the completed
-state. The package may lower both views into independent Presents in one peer `VisualTrack`.
-
-Triggered sequencing has the following exact law:
-
-1. preserve authored/source occurrence order and stable occurrence identity;
-2. never sort triggers by physical time;
-3. require at least one trigger, one explicit outer window and one explicit stage-terminal point;
-4. after frame quantization, require `B <= p1 < p2 < ... < pN < T <= O`;
-5. reject reversed triggers, equal-frame collisions, triggers outside the outer window and a last
- trigger at or beyond `T` atomically;
-6. zip an ordered item list to trigger occurrences only when the component contract explicitly says
- so and requires equal cardinality;
-7. derive item/stage identities from authored item and occurrence identities, never array indexes;
-8. keep cumulative overlap or successor handoff as declared package semantics, never Runtime choice.
-
-A package may instead give every child item its own single-occurrence Moment edge. Both author
-surfaces lower to the same ordered sibling candidates; positional zipping is not an implicit global
-behavior. Triggered stages are internal deterministic lowering facts unless another component has a
-real need to consume them. Composition still receives only ordinary Presents.
-
-## 5. Projection validation and negative intervals
-
-For each expanded occurrence, projection executes in this order:
-
-1. evaluate raw start and raw end in the exact rational ProgramSpace;
-2. reject missing, non-finite or type-invalid point expressions;
-3. if raw end precedes raw start, reject a reversed window before clipping;
-4. if raw end equals raw start, reject a zero-length render window;
-5. intersect the forward raw window with `[program.start, program.end)`;
-6. deterministically quantize it to a half-open integer frame span;
-7. reject an empty result or a result shorter than one frame;
-8. return only `{ startFrame, endFrameExclusive }` values wholly inside ProgramSpace.
-
-Clipping to ProgramSpace is the mathematical intersection with the only observable render domain,
-not guessed timing. A negative intermediate point is therefore allowed:
-
-```text
-cue = 1s
-[cue - 3s, cue + 2s] = [-2s, 3s]
-intersection [0s, 3s]
-```
-
-A raw reversed interval must never be hidden by clipping, sorting or absolute value. An interval
-entirely outside ProgramSpace becomes empty and fails; it is not silently omitted. Renderers receive
-only already validated frame spans and perform no temporal repair.
-
-A crossed Selection anchor pair is not by itself a projection failure. Identity projection
-`[selection.start, selection.end]` fails when it produces a reversed window, while
-`[selection.start, program.end]` may remain valid because it does not consume the crossed end point.
-
-## 6. Intrinsic duration and occupancy
-
-Every presented product either has a verified intrinsic duration or it does not. Unknown duration
-must not masquerade as either case; media inspection must establish it before an intrinsic-duration
-occupancy policy can execute.
-
-Durationless products such as Text, static images, Comment Stickers and ordinary Ranking UI occupy
-their complete final projected window. Local enter, exit and loop animation is presentation inside
-that window, not synthetic material duration.
-
-An intrinsic-duration product receives a final allowed window and an explicit playback policy. The
-conceptual visual policies are:
-
-```ts
-type IntrinsicPlayback =
- | { readonly mode: "once"; readonly align: "start" | "end" }
- | { readonly mode: "hold"; readonly align: "start" | "end" }
- | { readonly mode: "loop"; readonly align: "start" | "end" }
- | { readonly mode: "stretch" };
-```
-
-`once` replaces the ambiguous old word `finish`. End alignment never means reverse playback. It
-means the forward-playing source ends at the projected window's end. If the source is longer than
-the window, start alignment selects the source head and end alignment selects the source tail.
-
-| Policy | Source shorter than window | Source longer than window |
-|---|---|---|
-| `once/start` | play from window start, then disappear | play source head until window end |
-| `once/end` | appear late and end at window end | play source tail and end at window end |
-| `hold/start` | play, then hold last frame | play source head until window end |
-| `hold/end` | hold first frame, then play to window end | play source tail to window end |
-| `loop/start` | repeat from source head phase | clip after as many loops as fit |
-| `loop/end` | choose loop phase so source tail meets window end | same tail-phase law |
-| `stretch` | slow all source time into the window | accelerate all source time into the window |
-
-Audio must define its own truthful vocabulary over the same schedule. It cannot hold one PCM sample
-as if it were a video frame. Outside one-shot audio is silence, padding is explicit, looping is
-looping and stretching must enforce authored speed bounds. Static images and durationless UI do not
-accept meaningless intrinsic playback modes.
-
-Source trim and source in/out points, when added, operate before occupancy to define the effective
-intrinsic source interval. Spatial contain/cover and Text flow/shrink operate after temporal
-occupancy and are unrelated.
-
-## 7. Package ownership
-
-The former umbrella `contracts` package is intentionally absent. Nominal shared meanings are owned
-by focused modules such as Narrative, ProgramSpace, Media and Composition; Core has no registry of
-video-domain unions.
-
-The temporal algebra in this document is implemented by the focused video-domain package
-`@narratage/temporal`. It owns the projection types, validators and pure
-frame-window projection. It may consume Narrative point references, SemanticMap location and
-ProgramSpace, but it must not know Text, Media Track, Ranking, Comment Sticker, Film, HyperFrames,
-Runtime or a Provider.
-
-Official packages then use the same implementation:
-
-```text
-Text ───────────────┐
-Media Track ────────┤
-Deck Track ─────────┤
-Ranking ────────────┼─> @narratage/temporal ─> frame windows
-Comment Sticker ────┤
-Screen Overlay ─────┤
-Audio Track ─────────┘
-```
-
-Caption's component-owned timed Cue projection may enter the final frame-window stage directly.
-Speech Spine establishes ProgramSpace and contiguous speech takes. Other Tracks may address one
-whole Narrative Segment through its two structural anchors directly, without pretending it is an
-authored Selection.
-
-## 8. Migration consequences
-
-The current `typography-track` Surface consumes the shared temporal package for Program, Selection
-and Moment bindings, explicit point expressions and `one`/`each` expansion. It does not maintain a
-second timing enum. Media Track consumes the same package for Item and Sequence timing.
-
-The executable acceptance coverage includes:
-
-- Selection and Moment `one` success and cardinality failure;
-- Selection and Moment `each`, stable occurrence identity and source-order preservation;
-- rejection of occurrence-invariant `each` projections with multiple occurrences;
-- all common point-expression combinations, including Program boundaries and absolute time;
-- crossed source anchors consumed and ignored by different projections;
-- negative offsets, ProgramSpace intersection, entirely out-of-range windows, reversed windows,
- zero windows and the one-frame minimum;
-- rational frame rates and deterministic half-open frame quantization;
-- atomic failure when one of several `each` occurrences is invalid;
-- package-owned disjoint, independent and sequence relations remaining separate from expansion;
-- cumulative and successor-handoff trigger sequences, including explicit terminal boundaries,
- equal-frame collisions, non-monotonic source order, stable identities and optional explicit
- initial state;
-- durationless products and shorter/equal/longer intrinsic products under every supported playback
- and alignment policy;
-- explicit audio behavior rather than reuse of visual `hold` semantics.
-
-## 9. Ranking specialization
-
-The complete Ranking audit, component split, group schedule and executable migration are specified in
-[`ranking-track.md`](./ranking-track.md). Ranking now validates that the common temporal algebra can
-support cumulative state, exclusive current stages and a final settled suffix without adding
-Ranking or staging meaning to Core or the public Track contract.
-
-## 10. Migration state
-
-Comment Sticker content, layout and presentation are implemented as an independent author package;
-it shares only the common Temporal, Spatial and terminal VisualTrack vocabulary. No currently
-specified Track family remains blocked on this common authoring model.
-
-For every currently specified official Track package, the implemented acceptance matrix proves
-Text, Caption, Media, Deck, Ranking and Screen Overlay lower to peer VisualTrack values, while
-generic Audio and optional Media/Ranking sound lower to peer AudioTrack values without changing
-Core or Composition.
-
-Generic Audio Track and self-contained Screen Overlay are now specified separately in
-[`audio-track.md`](./audio-track.md) and [`screen-overlay.md`](./screen-overlay.md). Audio reuses the
-same temporal algebra but has truthful sample occupancy rather than visual hold semantics. Screen
-Overlay is a full-canvas peer contribution, not a cross-Track adjustment layer.
-
-Unified Item/Sequence Media authoring, all motion channels, internal handoffs, explicit audio
-projection and Speech Spine's restricted reuse are now specified in
-[`media-track.md`](./media-track.md). The old depth-stack Deck is an independent higher-order Track
-component specified in [`deck-track.md`](./deck-track.md). B-roll is an editorial Recipe/usage
-rather than another public Track family.
diff --git a/docs/guide/spec/track-composition.md b/docs/guide/spec/track-composition.md
deleted file mode 100644
index 967df71d..00000000
--- a/docs/guide/spec/track-composition.md
+++ /dev/null
@@ -1,141 +0,0 @@
----
-title: SVML Track and Composition
-description: Executable repository-internal `@1` compatibility waist for Track and Composition.
----
-
-# SVML Track and Composition
-
-Status: executable repository-internal `@1` compatibility waist; all gates in
-[`track-expressiveness.md`](./track-expressiveness.md) pass, publication remains separate.
-
-## Law
-
-Except for the shared ProgramSpace and canvas geometry, every audiovisual contribution entering a
-Composition is a peer Track. Composition validates, orders and combines Tracks. It does not know
-Caption, Speech, Media, Text, Seedance or any author-package family.
-
-```text
-Speech visual ───────────────> VisualTrack ─┐
-Media ───────────────────────> VisualTrack ─┤
-TimedCaption + style/cues ───> VisualTrack ─┤
-Text + style ────────────────> VisualTrack ─┼─> Composition
-Vignette / overlay ──────────> VisualTrack ─┤
-Speech audio / BGM / SFX ────> AudioTrack ──┘
-```
-
-## ProgramSpace
-
-ProgramSpace owns duration and exact rational frame rate. Its duration must end on an integer frame
-boundary. A Track does not copy a ProgramSpace digest into every value. The graph connects the Track
-and ProgramSpace explicitly wherever their relationship is validated; the validator rejects frame
-windows outside that supplied space.
-
-Canvas width, height and clear color are structural Composition facts. No video Track defines
-duration or becomes a privileged Base.
-
-## VisualTrack
-
-A VisualTrack is one self-contained render contribution owned by an author package. It explicitly binds the
-[`svml.visual-ir@1`](./visual-ir.md) terminal language and owns zero or
-more frame-exact `VisualPresent` values; every Present owns its own absolute `(order, tieBreak)`
-stacking key and one self-contained, code-free element tree made from box, text, ordinary media and
-typed compositable-Surface primitives. Parent references are Present-local. Media and exact font
-faces enter through content-addressed Artifact references rather than CSS URLs or environment font
-names. Every terminal text element carries a non-empty exact Font stack.
-Its ProgramSpace identity is intrinsic because the Track cannot be interpreted without a clock.
-Which Records produced it is Graph/Derivation truth and is deliberately absent from the Track.
-
-Composition flattens Presents across every Track before ordering them. One Media package may
-therefore own a board at z=30 and an icon at z=80 while a Text Present from another Track sits at
-z=50. A z change over time is represented by two non-overlapping Presents with different stacking
-keys. Track ownership never creates a render stacking context or a provenance chain.
-
-Every element may optionally carry frame-exact local keyframes over opacity, transform, filter or
-clip-path. Those keyframes operate only on that Present's element tree. A Media package may lower a
-pair transition into complementary animations on two of its own Presents without teaching
-Composition the name or semantics of that transition.
-
-VisualTrack deliberately has no:
-
-- family or source-kind field;
-- sibling Track or Present reference;
-- arbitrary script or global CSS selector;
-- backdrop filtering or blend mode that samples lower Tracks;
-- cross-Track mask, intersection or transition input.
-
-A local filter, crop, transform, animation or transition is valid when it operates only on materials
-explicitly owned by the authoring package and lowers to self-contained Presents. An effect that needs
-two raw materials must receive both explicitly before final Track lowering. No Present may sample
-the accumulated lower composite.
-
-## AudioTrack
-
-An AudioTrack owns exact 48 kHz sample-domain clips, typed canonical-WAV Artifacts, source interval,
-loop phase, pitch-preserving playback rate, target interval, gain and sample-exact fades. It has no
-visual stacking key, inert bus label or special relationship to a Speech Track. Speech audio, music,
-source audio and sound effects use the same contract.
-
-## Composition
-
-Composition contains one canvas and a set of VisualTrack/AudioTrack values. ProgramSpace is a peer
-input edge to Film and final-render Operations rather than repeated lineage metadata. Composition:
-
-1. validates every Track against the ProgramSpace supplied by an explicit graph edge;
-2. rejects duplicate Track identities and duplicate visual stacking keys;
-3. flattens every VisualPresent and mounts it by absolute `(order, tieBreak)`;
-4. mixes AudioTrack clips in ProgramSpace;
-5. produces a renderer-neutral input for an explicitly selected final-render package.
-
-Composition must not switch on an author-domain family or grant a Track special access to accumulated
-lower pixels. A global shatter, adjustment layer, Base FX lane or B-roll underlay transition is
-therefore not part of this version. No dedicated Base FX placeholder is reserved.
-
-TimedCaptionProjection remains an intermediate semantic value. Caption grouping, style and cue
-lowering must finish before the caption enters Composition as an ordinary VisualTrack.
-
-## Final-render boundary
-
-`@narratage/hyperframes` is the current reference compiler for Composition and its one versioned
-SVML Visual IR. It never
-switches on Speech, Caption, Media or author-package identity. It deterministically emits a
-content-addressed `HyperframesDocument` whose HTML interleaves VisualPresents by absolute stacking
-key. It must not mount an authoring Track as one isolated visual wrapper. AudioTrack compilation
-and final mux are separate media operations over the same ProgramSpace; HyperFrames is a silent
-visual target.
-
-This is not an exclusive boundary. A future `render-remotion` or API-backed render package may
-consume the same Composition and ProgramSpace, compile another immutable renderer document and
-return the same media contract. Author Source selects that path by importing its package; Core and
-Runtime do not guess a renderer. The current examples select `render-hyperframes` explicitly.
-
-The document MUST bind exact rational frame rate, positive integer frame count, canvas dimensions,
-Artifact set and generated HTML. ProgramSpace relationship remains visible in the Producer's input
-edges and Derivation rather than being copied into document content. Its complete visual frame domain is the
-half-open interval `[0, frameCount)`. Rendering frame `n` must depend only on this immutable document,
-its exact Artifacts, the locked renderer implementation and `n`; it must not require sequentially
-evaluating frames `0..n-1`. Stateful author effects must therefore be materialized into a typed
-Surface before entering this boundary.
-
-Frame partitioning is Runtime policy. A conforming Endpoint may split the domain into any set of
-non-overlapping half-open spans whose ordered union is `[0, frameCount)`, evaluate them locally or
-remotely, retry individual spans, then assemble visual chunks and mux audio once. Those spans are
-not author declarations, Core Operations or Track identities. The fulfilled render Product must
-repeat the exact document/frame-domain/canvas bindings; affinity rejects a result assembled for a
-different domain.
-
-The compiled document contains `svml-artifact://` placeholders, not local paths, API URLs or signed
-URLs. Resolving those placeholders is a Runtime/Provider action immediately before rendering. Thus
-one compiled document can be rendered locally or remotely without changing author intent or its
-compiled identity. Exact text becomes generated `@font-face` rules with font synthesis disabled.
-Typed Surfaces preserve declared dimensions, color space, alpha mode and still/frame timing across
-the same Artifact boundary; they are not inferred from filename extensions. A renderer validates
-those byte facts before staging or declines the document. Its generic Need Receipt binds the locked
-Endpoint implementation, and renderer-specific evidence in receipt-covered metadata binds the exact
-browser or immutable remote deployment used for layout.
-
-`@narratage/film` implements package-level arbitrary-arity assembly as a finite immutable TrackSet
-fold followed by ordinary Composition. Rendering is a separate explicitly imported downstream
-Fragment. Core receives only fixed-port, single-result Operations, and targeting an intermediate
-Track or Composition does not demand a renderer. The Film Surface must lower into the implemented
-Fragment without adding Track families to Core or making one Track depend on accumulated sibling
-pixels.
diff --git a/docs/guide/spec/track-expressiveness.md b/docs/guide/spec/track-expressiveness.md
deleted file mode 100644
index 4cef7571..00000000
--- a/docs/guide/spec/track-expressiveness.md
+++ /dev/null
@@ -1,210 +0,0 @@
----
-title: SVML Track Expressiveness Gate
-description: all E1–E8 gates execute; repository-internal `svml.visual-track@1` waist frozen pre-publication.
----
-
-# SVML Track Expressiveness Gate
-
-Status: all E1–E8 gates execute; repository-internal `svml.visual-track@1` waist frozen pre-publication.
-
-## Purpose
-
-One public Track contract must solve final audiovisual composition without pretending that every
-author package shares one authoring model. Text, Caption, Media, Ranking and future packages own
-their Programs and may revise those Programs independently. They may enter Composition only after
-lowering their package-specific meanings into self-contained, frame-exact contributions.
-
-This gate answers one question:
-
-> Can production-used text, media-box, caption, stacking and local-effect behavior lower into the
-> same Track waist without adding package-family fields to Track, Composition or Core?
-
-Passing this gate does not make one universal Track Program. The six Track Program axes—content,
-temporal source, window projection, spatial source, occupancy and presentation—remain a separation
-discipline for package authors, not six mandatory public fields. The shared authoring laws and the
-current temporal design are specified in [`track-authoring.md`](./track-authoring.md).
-
-## Three levels that must not collapse
-
-```text
-Package Program Resolved package facts Public Track
-─────────────────────── ──────────────────────────── ─────────────────────
-TextLayout / CaptionMode boxes, glyph runs, windows Presents
-Media sampling policy -> source-time and crop mapping -> exact frame spans
-effect recipe local animations absolute stacking
-semantic/spatial locator final placement self-contained elements
-```
-
-The Package Program is editable author truth. Resolved package facts are compiler evidence. The
-public Track is a terminal audiovisual contribution. Composition must never reconstruct the first
-two levels from the third.
-
-"Flat Track" means there is no Track nesting or implicit Track stacking context. A Present may
-still own an internal element tree. One authoring Track may emit several Presents at unrelated
-absolute stack positions so that peer Tracks can interleave between them.
-
-## Required witnesses
-
-### E1 — Text three-box lowering
-
-The official browser-oriented reference lowering must represent independently:
-
-1. an outer placement frame;
-2. an inner layout box controlling wrapping, alignment, direction and overflow;
-3. paint targets for the complete frame, shrink-wrapped content, each line or each word.
-
-Paint overflow such as glow or shadow must not become a line-measurement input merely because it is
-visible outside the content box. A package may lower a paint target by attaching paint to a
-different owned element; the public Track must not acquire a `textBackgroundTarget` field.
-
-### E2 — Caption range/cue/content lowering
-
-A Caption package must be able to lower a hard range boundary, a cue-level alignment/animation
-container, shrink-wrapped rows and independently styled or animated words. Dual-font and per-word
-paint are package facts. Composition sees only ordinary elements and Presents.
-
-### E3 — Media two-frame lowering
-
-A media package must be able to lower one resolved Placement Frame containing two independent
-samples of the same Artifact. Each sample derives its own Content Frame according to
-[`spatial-layout.md`](./spatial-layout.md):
-
-- a backdrop sample with its own cover/zoom/blur/tone treatment;
-- a foreground sample with its own contain/cover/fit-height and focal alignment.
-
-The public Track must not acquire B-roll, focal-preset or blurred-background fields. Artifact
-dependency collection must deduplicate the shared content digest.
-
-### E4 — Independent absolute stacking
-
-One package may emit a board at one absolute stack position and icons or labels at other positions.
-A Present from another Track must be able to appear between them. Track identity groups one
-self-contained render contribution; provenance lives in the Graph/Derivation and Track never forms
-a render stacking context.
-
-### E5 — Local motion and handoff
-
-Frame-exact opacity, transform, filter and clip animation may operate on elements owned by one
-Present. A package that owns both materials in a handoff may lower complementary Presents. No
-ordinary Track may sample the accumulated pixels below it or mutate a sibling Track.
-
-The Media package must additionally prove that entry, sustain, exit, per-layer sampling motion and
-Sequence pair handoffs remain separate channels with a fixed composition order. The independent
-Depth-Stack Deck package must prove that its collection-state reflow composes with Card-local motion
-without adding Deck meaning to the terminal Track. The complete models are specified in
-[`media-track.md`](./media-track.md) and [`deck-track.md`](./deck-track.md).
-
-### E6 — Materialized visual fallback
-
-A component whose deterministic visual result cannot be represented by the reference element tree
-must be able to materialize its owned result and contribute it as a typed Surface. For transparent
-interleaving, the final public contract must bind enough surface information to distinguish a
-compositable alpha-bearing visual from an opaque video. Merely naming a `.webm` Artifact does not
-prove this property.
-
-### E7 — Generic audio lowering
-
-The official Audio package must lower arbitrary explicit normalized sources, source trim, one-shot,
-loop, bounded pitch-preserving stretch, start/end alignment, gain and fades into the same peer
-`AudioTrack` waist. Overlapping clips mix without priority clipping or a privileged Base lane.
-Container stream selection, loudness processing and Provider placement remain explicit upstream or
-downstream graph work. The complete boundary is specified in
-[`audio-track.md`](./audio-track.md).
-
-### E8 — Self-contained screen overlay
-
-A full-canvas effect must lower to ordinary absolute-stack Presents made from owned Visual IR
-elements or an owned alpha-bearing Surface. It cannot read the accumulated lower composite, reserve
-the highest z-index or introduce a post-composition phase. Source-free blur, color-adjust and real
-zoom-blur therefore fail closed; their honest form consumes explicit media or Surface input outside
-the Screen Overlay package. See [`screen-overlay.md`](./screen-overlay.md).
-
-## Old-system attack matrix
-
-| Historical behavior | Package-owned truth | Required terminal witness |
-|---|---|---|
-| Text placement/layout/paint boxes | Text Program | nested owned elements |
-| frame/content/line/word background | Text Program | paint attached to the selected owned level |
-| caption range/cue/content boxes | Caption Program | nested owned elements and word-local animation |
-| dual fonts, CJK, emoji and writing direction | Text/Caption Program | text elements plus locked font/layout dependencies |
-| media content box plus foreground/backdrop sampling | media package | sibling media elements referencing one Artifact |
-| focal crop and fit-height | media package | deterministic sampling/placement result |
-| full-frame B-roll, corner GIF and lower media card | Media Item + Recipes | the same Item lowering against different Frames |
-| explicit media replacement group | Media Sequence | coordinated owned outgoing/incoming Presents |
-| visible card history and depth reflow | Depth-Stack Deck Track | package-owned collection state over ordinary Presents |
-| Ranking board and icons at unrelated z | Ranking Program | several absolute-stack Presents |
-| Media enter/exit and pair transition | Media Program | local frame-exact keyframes over owned Presents |
-| arbitrary self-contained visual | component/provider | materialized compositable Surface Artifact |
-| music/SFX/additional voice placement | Audio Program | ordinary frame-exact AudioTrack clips |
-| flash/vignette/grain/veil | Screen Overlay Program | owned full-canvas Presents or alpha Surface |
-| blur/color transform of a picture | explicit-input media effect | transformed owned material, never backdrop sampling |
-
-## Current executable evidence
-
-`packages/hyperframes/test/track-expressiveness.test.ts` constructs package-private lowering
-witnesses rather than adding Text, Caption, Media or Ranking fields to public contracts. Existing
-package tests separately prove ordinary Text, Caption and Media Programs lower through the same
-Track and Composition path.
-
-The current `VisualTrack / VisualPresent / VisualElement` candidate explicitly binds
-`svml.visual-ir@1` and can encode the structural three-box, two-box, word-local and
-absolute-stack cases. Two generic terminal facts are now
-executable rather than package-family patches:
-
-1. `FontArtifactRef` binds each exact fallback face, weight and style to one or more Blob sources.
- Unicode-range shards remain one logical face; exact text lowering emits content-addressed
- `@font-face` rules for every source and disables font synthesis;
-2. `CompositableSurfaceRef` binds dimensions, sRGB color space, opaque/straight alpha semantics and
- still/frame timing to Blob bytes. Animated Surfaces must exactly share the containing Present's
- frame count and ProgramSpace rate.
-
-`packages/hyperframes/test/browser-visual.test.ts` additionally drives real Hyperframes browser
-frames: installed content-addressed open fonts paint Latin, CJK and emoji/symbol glyphs through an ordered fallback
-stack; Text proves Point/Area/Path layout, all Box targets, rich runs, horizontal/vertical flow,
-clip/ellipsis/bounded shrink, ordered Paint, local Mask and forward/reverse selector clocks; Fine
-Caption proves multiline wrapping, full glyph Paint, every glyph karaoke mode and joined trail Pill
-line geometry; DepthStack Deck proves explicit old/new collection-state reflow, exact labels and
-byte-identical one-worker/partitioned frames; all eleven official Screen Overlay components preserve identical pixels under sequential and parallel frame rendering; and a
-50% straight-alpha PNG composites over a blue Track at the expected pixel values. This opt-in host
-test is run with `SVML_BROWSER_TESTS=1`; it proves the current local renderer path, not every future
-hosted Runtime.
-
-The IR style vocabulary is closed rather than arbitrary CSS: components cannot add an unknown
-property, environment-dependent value or alternate browser language without changing the protocol
-version. Every terminal text element now carries a non-empty exact Font Artifact stack; system-font
-fallback and a prototype family are invalid. The Runtime binds its locked implementation in the
-generic Receipt. HyperFrames additionally records the exact local browser digest or configured
-remote renderer-deployment digest in receipt-covered metadata. Shared ffprobe-based Surface-byte
-verification proves declared dimensions, timing, SDR/sRGB and alpha facts before local staging;
-Endpoints without this ability decline Surface-bearing documents.
-
-E1–E5, E7 and E8 now have their official package witnesses. `@narratage/typography-track` proves the
-complete three-box/Text terminal model without public Text fields; Caption and Media prove their
-independent package models. `@narratage/audio-track` proves the exact
-48 kHz source/target mapping, overlap law, bounded stretch and one shared local/remote
-`AudioProgramPlan`; `@narratage/screen-overlay` proves only self-contained owned pixels, explicit
-seeds, flat stacking and fail-closed rejection of lower-composite effects;
-`@narratage/deck-track` proves pure absolute-frame collection reflow over the union of old/new
-visible Cards while reusing Card-local Media lowering. `@narratage/ranking` supplies the real E4
-witness: board, stage and independently stacked Items interleave with an unrelated peer Track,
-while all four Ranking components retain frame-pure progressive state under partitioned browser
-rendering. E6 is complete through the typed path, independently installed non-native Text witness,
-byte verifier and fail-closed Endpoint support check.
-
-## Freeze criteria
-
-The repository-internal Track v1 waist satisfies all of the following:
-
-1. E1–E8 have executable structural, HTML-lowering and audio-plan witnesses;
-2. official Text/Caption lowerers either bind exact font Artifacts or explicitly materialize their
- glyph result, and the renderer implementation used for layout is receipt-bound;
-3. E6 has a typed compositable-Surface path whose Runtime validates the declared media facts and
- records the renderer implementation;
-4. installing a new Track package requires no Core, Film, Composition or family registry change;
-5. public Track data contains no Caption, Text, Media, Ranking or provider discriminator;
-6. a package can revise its Program schema while preserving or versioning only its own lowerer;
-7. unsupported cross-Track sampling fails closed.
-
-The implementation is therefore frozen for subsequent repository package work. Public npm
-publication, untrusted-code isolation and final release versioning remain separate release gates;
-they do not reopen the video meaning of this waist.
diff --git a/docs/guide/spec/typography-track.md b/docs/guide/spec/typography-track.md
deleted file mode 100644
index deb7c2d6..00000000
--- a/docs/guide/spec/typography-track.md
+++ /dev/null
@@ -1,443 +0,0 @@
----
-title: Typography Track Authoring and Expressiveness
-description: Implemented pre-release authority for the official two-dimensional Text package.
----
-
-# Typography Track Authoring and Expressiveness
-
-Status: implemented pre-release authority for the official two-dimensional Text package. The
-complete declared model and browser evidence execute through the frozen repository-internal
-Track/Visual IR waist; the author Surface itself is not yet a published ABI.
-
-## 1. Conclusion
-
-The complete Text model is the orthogonal product of:
-
-```text
-authored document
-× temporal binding
-× spatial form
-× text flow
-× exact typography
-× ordered Paint layers
-× deterministic local motion
-```
-
-An ordinary transparent title, one background around the complete phrase, one background per
-rendered line, isolated word Pills, connected highlighter geometry and a per-grapheme typewriter are
-different values in this model. They are not different renderer components and do not require Core,
-Film or Composition branches.
-
-"Complete" does not mean putting every effect ever made into one flat object. It means:
-
-1. every ordinary two-dimensional editor Text behavior has a typed place in this model;
-2. new Styles and Recipes do not change shared protocols;
-3. capabilities with genuinely different inputs or physics use explicit sibling components;
-4. any self-contained visual that exceeds the terminal element language can materialize a typed
- compositable Surface without changing Core.
-
-The official Text package owns this author meaning. SVS only provides named immutable Recipe
-parameters. Visual IR remains the shared terminal video language. HyperFrames is one renderer of
-that language, not the definition of Text.
-
-## 2. Ownership and package boundary
-
-The ordinary two-dimensional package remains one package:
-
-```text
-@narratage/typography-track
- Point Text
- Area Text
- Path Text
- Text Style / Inline Style declarations
- Text Program validation
- temporal/spatial resolution adapters
- VisualTrack lowering
-```
-
-Point, Area and Path are separate author components because they assert different geometry. They
-are not a `mode` field which makes ports appear and disappear. They may still lower into one tagged
-union and coexist as peer items in one Text Track.
-
-Two capabilities have materially different inputs and stay out of ordinary Text:
-
-- a Text Mask component must explicitly consume both the text shape and the media/graphic it masks;
- it may not sample whichever Track happens to be below it;
-- true 3D Text owns extrusion, material, light and camera semantics and normally materializes an
- alpha-bearing Surface.
-
-Those may be separate packages or separately imported components in a later official distribution.
-Package count is a distribution decision; their graph contracts must remain separate. Neither case
-adds a branch to Core.
-
-The official local Text Mask witness deliberately accepts only one exact single-line Area Text
-shape and one explicit still `CompositableSurface`. Rich runs, multiline/Path flow, sequence
-animation and timed materials fail closed and use the same materialized-Surface escape route as 3D
-Text. This boundary avoids pretending that browser `foreignObject` masking is portable or exact.
-
-## 3. Authored Text Document
-
-A Text item consumes a bounded rich document rather than an HTML fragment or arbitrary tree:
-
-```ts
-type TextDocument = {
- readonly paragraphs: readonly TextParagraph[];
-};
-
-type TextParagraph = {
- readonly id: string;
- readonly runs: readonly TextInline[];
- readonly style?: TextParagraphStyleRef;
-};
-
-type TextInline =
- | { readonly kind: "text"; readonly text: string; readonly style?: TextInlineStyleRef;
- readonly language?: string; readonly direction?: "auto" | "ltr" | "rtl" }
- | { readonly kind: "break" };
-```
-
-The exact public names may change during implementation, but the laws do not:
-
-1. text, whitespace, punctuation and explicit breaks preserve author order exactly;
-2. a plain string is shorthand for one paragraph with one run;
-3. a Span changes only explicitly inheritable typography/Paint, never timing or item placement;
-4. source text is segmented only when a selected layout/Paint/motion operation needs units;
-5. segmentation is Unicode- and locale-aware and implementation-bound;
-6. no Text package uses an LLM to rewrite, split or decorate authored content;
-7. `uppercase`, small caps and similar transformations are presentation and never mutate the
- authored document value.
-
-The deterministic unit hierarchy is:
-
-```text
-document > paragraph > explicit run > rendered line > Unicode word > grapheme cluster
-```
-
-`rendered line` exists only after exact fonts and final geometry are known. A grapheme cluster is
-the smallest ordinary author-visible animation/decoration unit. A shaped glyph is not always in
-one-to-one correspondence with a Unicode character or grapheme, so the public author model must not
-pretend otherwise.
-
-## 4. Three spatial forms
-
-Temporal source and window projection remain governed by
-[`track-authoring.md`](./track-authoring.md). Text does not invent `full`, `from`, `until` or another
-timing system.
-
-Spatial placement is independent from Text flow. The shared model in
-[`spatial-layout.md`](./spatial-layout.md) provides resolved point/frame/path geometry; Text then
-interprets that geometry through one of three forms.
-
-### 4.1 Point Text
-
-Point Text is intrinsic text anchored at a resolved point. It normally hugs its content, does not
-soft-wrap and uses an explicit inline/block anchor. Manual paragraph breaks remain valid. Ordinary
-titles, labels and stickers are Point Text.
-
-### 4.2 Area Text
-
-Area Text flows inside a resolved frame. It explicitly chooses:
-
-- inline and block sizing: `hug | fixed` where meaningful;
-- padding on four logical edges;
-- inline alignment: `start | center | end | justify`;
-- block alignment: `start | center | end`;
-- wrap: `none | word | grapheme`;
-- overflow: `visible | clip | ellipsis | shrink`;
-- optional maximum lines for the truncating/shrinking policies;
-- writing direction and writing mode;
-- optional columns and column gap;
-- clipping to the placement frame;
-- metric edge: line box, cap height or ink bounds.
-
-Overflow is honest author intent. `ellipsis` may omit visible authored text and therefore must be
-explicit. `shrink` must declare a minimum scale and fail if the complete text still cannot fit; it
-must not silently cross that bound. This differs from Caption, which must never clip or discard the
-author's display Atoms.
-
-### 4.3 Path Text
-
-Path Text consumes an explicit owned vector path plus Text Document and Style. It owns path side,
-orientation, start/end margins, alignment, reversal and overflow. Animating a margin moves text
-along the path; animating the path changes the path. It never discovers or samples a shape in
-another Track.
-
-Path Text requires either a generic vector/path terminal primitive or an owned materialized
-Surface. It cannot be claimed by serializing an untyped SVG/HTML string.
-
-## 5. Typography
-
-Typography is separate from Paint and layout but participates in layout measurement. It includes:
-
-- an exact ordered `FontStackRef`, with environment family names only as an explicit prototype path;
-- size, weight and normal/italic/oblique style;
-- variable-font axes;
-- OpenType feature choices and font synthesis policy;
-- kerning, tracking, word spacing and line height;
-- language, direction and writing mode;
-- baseline shift, tab width, indentation and paragraph spacing;
-- underline/overline/strike geometry where it follows glyph metrics;
-- text transform, small caps, superscript and subscript presentation;
-- CJK-specific spacing/compression controls when supported by the selected layout implementation.
-
-Exact font bytes remain explicit author graph inputs. They are not names hidden in a Recipe and are
-not selected by the Runtime. Font shaping/layout implementation identity is bound by the accepting
-derivation/render receipt; Core still knows nothing about fonts.
-
-## 6. Ordered Paint model
-
-Paint is an ordered list, not one mutually exclusive set of CSS shortcuts. The Text Program owns
-typed layers such as:
-
-```ts
-type TextPaintLayer =
- | { readonly kind: "fill"; readonly paint: ColorPaint }
- | { readonly kind: "stroke"; readonly paint: ColorPaint; readonly widthPx: number;
- readonly placement: "inside" | "center" | "outside" }
- | { readonly kind: "shadow"; readonly paint: ColorPaint; readonly offset: Vector;
- readonly blurPx: number; readonly spreadPx: number }
- | { readonly kind: "glow"; readonly paint: ColorPaint; readonly blurPx: number;
- readonly spreadPx: number }
- | { readonly kind: "box"; readonly target: TextPaintTarget;
- readonly continuity: "isolated" | "joined"; readonly decoration: BoxDecoration };
-```
-
-`VisualColorPaint` supports solid, linear-gradient and radial-gradient Paint with an ordered list of color
-and opacity stops. Repeating a Stroke or Shadow is legal and order remains semantic. A renderer may
-lower an outside stroke through duplicated glyph layers, but that is implementation, not an author
-flag whose meaning changes by browser.
-
-Box Paint has these targets:
-
-```text
-frame | content | paragraph | line | run | word | grapheme
-```
-
-Its decoration independently owns fill, border, padding, corner radii and shadows. `isolated`
-creates one decorated unit per target. `joined` unions adjacent selected fragments within each
-rendered line and gives the line fragment honest end caps. Invalid combinations, such as continuity
-on `frame`, fail during Text validation.
-
-This one rule expresses the common cases:
-
-| Look | Paint value |
-|---|---|
-| ordinary transparent text | no Box Paint layer |
-| one phrase capsule | `target=content` |
-| full placement panel | `target=frame` |
-| one box per rendered line | `target=line, continuity=isolated` |
-| one Pill per word | `target=word, continuity=isolated` |
-| connected marker/highlighter | `target=word, continuity=joined` |
-| letter tiles | `target=grapheme, continuity=isolated` |
-
-A callout tail is an explicitly authored attachment to one Box Paint layer, with side, offset, size
-and its own Paint. It does not silently force a background. Frosted glass is not ordinary Box
-Paint because it samples pixels behind the Track; it requires an explicitly owned media input or a
-materialized self-contained component.
-
-## 7. Motion and sequence selectors
-
-Motion is independent from Style so the same look can enter, loop and exit differently. One item
-may have:
-
-1. item-level finite keyframes over position, scale, rotation, skew, opacity, blur and clipping;
-2. path-margin animation for Path Text;
-3. one or more sequence animators over text units.
-
-A sequence animator is the product of:
-
-```text
-unit selector × selected property channels × keyframes × stagger/order
-```
-
-The unit selector may target paragraph, line, run, word or grapheme. It explicitly chooses range,
-forward/reverse order, delay/spread, number of cycles and any deterministic random seed. Property
-channels include local transform, opacity, blur and glyph Paint values. Typewriter is a grapheme
-reveal preset over this model, not string slicing. Rolls and crawls are ordinary item-level motion
-over Area Text.
-
-No motion name is a hidden callback. Presets such as `pop`, `spring`, `slide` or `wobble` compile to
-finite or analytically bounded package-owned motion values before terminal rendering. Two motion
-channels compose through separate owned wrappers or an explicit transform composition order; the
-last one does not overwrite the previous one.
-
-Animation lifetime is independent from Item visibility. A motion may finish before its Item window;
-its final authored state then holds until the Item disappears. A motion may also extend beyond the
-window; the Item window clips it without rewriting its timing. Sequence and Path Text motion follow
-the same rule. This avoids rejecting a perfectly meaningful fast title animation merely because the
-title remains visible, and avoids silently retiming a slow animation to fit a short window.
-
-## 8. Style declaration, SVS and readable authoring
-
-Declaration and use remain separate. SVS stays a generic scalar Recipe language; it must not learn
-Text Paint arrays or become a second renderer. A Text Style Surface reads named Recipes, exact Font
-references and repeated structured layer declarations, validates them, then emits one complete
-package-owned `TextStyle` Record.
-
-The intended author shape is:
-
-```svml
-<text:Style id="plain" recipe={studio.text.plain} font={fonts.inter}/>
-
-<text:Style id="word-pills" recipe={studio.text.word-pills} font={fonts.inter}>
- <text:Box target="word" continuity="isolated" recipe={studio.text.word-pill}/>
-</text:Style>
-
-<text:Style id="poster" recipe={studio.text.poster} font={fonts.inter}>
- <text:Stroke recipe={studio.text.poster-outline}/>
- <text:Stroke recipe={studio.text.poster-keyline}/>
- <text:Shadow recipe={studio.text.poster-shadow}/>
-</text:Style>
-
-<text:Track id="titles" space={film.space}>
- <text:Point id="hook" style={poster} during={story.selection.hook} placement={hook-point}>
- Nothing hidden in the Runtime.
- </text:Point>
-
- <text:Area id="explanation" style={word-pills} during={story.selection.body} placement={body-frame}>
- <text:P>Every <text:Span style={plain}>visible word</text:Span> stays authored.</text:P>
- </text:Area>
-</text:Track>
-```
-
-An item may alternatively consume one ordinary graph `Text` value:
-
-```svml
-<copy:Value id="headline">Nothing hidden in the Runtime.</copy:Value>
-
-<text:Track id="titles" space={film.space}>
- <text:Point id="hook" content={headline}
- style={poster} during="program" placement={hook-point}/>
-</text:Track>
-```
-
-`content={Text}` and inline body content are mutually exclusive. The former materializes one plain
-document run from the exact graph value; the latter owns the bounded rich document and may use
-`P`, `Span` and `Break`. This prevents a Frontend from copying runtime Text into hidden authored
-state while keeping rich typography explicitly owned by this package.
-
-This spelling executes through the current shared temporal/spatial Surface syntax. Its important
-properties are:
-
-- the author imports and selects the Typography package;
-- Styles are declared before use and are ordinary authored Records;
-- simple Style values remain compact;
-- repeated layers are repeated declarations, not numbered fields or encoded strings;
-- the Markup Surface, not SVS, interprets the Recipe vocabulary;
-- exact fonts and timing/placement facts travel on explicit graph edges;
-- the resulting Text Program contains no unresolved Recipe and no Runtime/provider choice.
-
-## 9. Compiled pipeline and data gates
-
-```text
-Markup Surface
- (authored rich document | graph Text input) + named Style + temporal binding + spatial binding
- │
- ▼
-Text Program package-owned author truth
- │
- ├── located temporal points ──> projected frame windows
- └── resolved spatial geometry
- │
- ▼
-Resolved Text Layout
- exact fonts + shaping + wrapping + fragment geometry + motion schedule
- │
- ▼
-VisualTrack
- code-free terminal elements/keyframes or typed owned Surface
- │
- ▼
-Composition -> selected final renderer
-```
-
-Graph edges carry every external dependency. Text values contain only intrinsic Text meaning. No
-upstream Record gains text, role, source, style, font or layout metadata merely because a later Text
-consumer needs it.
-
-The layout implementation may use a locked browser, HarfBuzz/Skia/Pango or another exact engine.
-It must bind the actual implementation and font bytes. Browser line boxes are not propagated as
-generic graph metadata; they are either deterministic terminal layout owned by the renderer receipt
-or package-owned resolved geometry used to materialize a Surface.
-
-## 10. Terminal Visual IR findings
-
-The frozen repository-internal `svml.visual-ir@1` carries the smallest additions proven by this
-migration:
-
-- `text-flow` owns the bounded rich document, exact typography, ordered Paint and sequence values;
-- `path-text` owns typed vector commands and exact path-flow facts;
-- the local `mask` primitive owns exactly one terminal mask source and one content root;
-- ordered repeated Paint and Unicode unit animation remain serialized data, never callbacks;
-- exact font Artifacts and typed `CompositableSurface` values cross the ordinary Artifact boundary.
-
-The HyperFrames reference compiler implements those primitives without exposing HTML, CSS, SVG or
-renderer scripts to author packages. A separately installed non-native Text fixture proves that
-extrusion/material/light/camera semantics can remain outside the waist and contribute only a typed
-alpha Surface. Unsupported rich masks use the same path.
-
-These remain video-terminal facts, never Core facts. Renderer receipts, Surface-byte validation,
-Deck/Ranking witnesses and the final compatibility audit now pass without adding Text meaning to
-the shared waist.
-
-## 11. Feedback into Fine Caption
-
-Text and Caption share typography implementation problems, not author truth.
-
-The following may be shared through a focused video-domain implementation library, provisionally
-`@narratage/typography`, with no Author Surface and no graph Type unless a real cross-package edge
-later requires one:
-
-- Unicode grapheme/word/bidi segmentation;
-- exact font-stack loading and shaping;
-- gradient stops, ordered strokes/shadows/glows and box decoration geometry;
-- line-fragment and joined-Pill geometry;
-- grapheme-safe typewriter and deterministic selector evaluation;
-- terminal rich-text lowering helpers.
-
-This does not create a universal Text Program. Caption keeps `CaptionDisplaySequence`, Cue, Atom,
-speech correspondence and karaoke timing. Text keeps paragraphs, rich runs, point/area/path layout
-and arbitrary sequence selectors.
-
-Concrete Caption improvements discovered by this audit are:
-
-1. allow repeated ordered Stroke/Shadow/Glow layers instead of the current one-of-each ceiling;
-2. allow arbitrary gradient stops and radial as well as two-stop linear gradients;
-3. reuse explicit inner/center/outer Stroke semantics rather than renderer-dependent outline
- behavior;
-4. reuse grapheme-safe text transform/reveal and exact shaping;
-5. keep joined active-Pill geometry on the shared line-fragment implementation.
-
-Caption must **not** inherit Text ellipsis, clipping, shrink-to-fit, path flow, arbitrary rich-run
-timing or selector-generated speech timing. It always displays the author's complete immutable
-Atoms and activates only from proven Atom windows.
-
-## 12. Migration order and acceptance gates
-
-Implementation replaced the earlier slice rather than accreting compatibility fields:
-
-1. **Implemented:** Text Document, Point/Area tagged geometry, Text Style and ordered Paint values;
-2. **Implemented:** exact fonts, frame/content/paragraph/line/run/word/grapheme Box Paint and rich-run static lowering;
-3. **Implemented:** shared Temporal Window Projection and explicit Point/Frame/Path Spatial inputs;
-4. **Implemented:** deterministic overflow, direction/writing mode and browser pixel evidence;
-5. **Implemented:** item motion and Unicode-aware sequence selectors;
-6. **Implemented:** Path Text plus an exact bounded local Text Mask and fail-closed materialization boundary;
-7. **Implemented:** the `CompositableSurface` escape route with a separately installed non-native Text package witness;
-8. **Implemented:** freeze the repository-internal Visual IR after every repository-wide gate passes.
-
-Acceptance requires browser/render evidence for at least:
-
-- transparent Point Text and fixed Area Text;
-- content, frame, multiline line, isolated word, joined word and grapheme boxes;
-- nested run styles across wrapping boundaries;
-- Latin, CJK, RTL, combining marks and color Emoji with exact fonts;
-- repeated outside/inside strokes, multiple shadows, linear/radial gradients and glow;
-- visible/clip/ellipsis/bounded-shrink overflow;
-- item enter/exit/loop and forward/reverse grapheme/word/line sequencing;
-- Path Text;
-- explicit Text Mask ownership and one materialized advanced fallback;
-- no modification to Core, Runtime, Film or an unrelated Track package.
-
-The declared Text package gates now pass. This establishes a complete pre-release Text package on
-the frozen repository-internal Track/Visual IR waist; npm publication and the Text author Surface's
-own public-version promise remain separate.
diff --git a/docs/guide/spec/visual-ir.md b/docs/guide/spec/visual-ir.md
deleted file mode 100644
index 92bb9da9..00000000
--- a/docs/guide/spec/visual-ir.md
+++ /dev/null
@@ -1,134 +0,0 @@
----
-title: SVML Visual IR
-description: implemented repository-internal `@1` compatibility waist; not yet published as an npm ABI.
----
-
-# SVML Visual IR
-
-Status: implemented repository-internal `@1` compatibility waist; not yet published as an npm ABI.
-
-## Position
-
-`svml.visual-ir@1` is the one code-free terminal visual language shared by the official
-SVML video stack. It is:
-
-- a public video-domain protocol;
-- not an author component;
-- not a Provider API;
-- not part of the domain-neutral Core.
-
-Text, Caption, Media, Ranking and third-party visual packages may own unrelated author Programs.
-Before entering a `Composition`, each package must lower its resolved result into a `VisualTrack`
-that explicitly names this IR. A renderer adapter validates and compiles that common language; it
-never learns the originating component family. `@narratage/hyperframes` is the first reference
-adapter, not the owner of the protocol.
-
-```text
-author Program shared terminal protocol final-render route
-───────────────────── ───────────────────────── ─────────────────────
-Text / Caption / Media ───> SVML Visual IR in Track ───┬─> HyperFrames
-custom visual producer ───> typed CompositableSurface ───┤
- ├─> future Remotion
- └─> future render API
-```
-
-Calling this a component "dialect" is misleading. There is one official target protocol for the
-current video stack. A component cannot import another set of CSS semantics and still claim to
-produce the same IR.
-
-## Contract
-
-Every `svml.visual-track@1` carries:
-
-```text
-visualIr = svml.visual-ir@1
-```
-
-The value is covered by the Record that carries the Track and then transitively by the Records for
-Composition and the selected renderer document. A Runtime cannot reinterpret an already compiled
-Track under a different visual language.
-
-The IR contains only:
-
-- frame-exact Presents with absolute stacking keys;
-- one Present-local tree of `box`, `text`, `image`, `video` or `surface` elements;
-- content-addressed media, font and Surface references;
-- a closed list of browser-oriented style declarations;
-- frame-addressed local keyframes over opacity, transform, filter and clip;
-- safe data/ARIA attributes.
-
-It contains no component family, author recipe, provider name, selector, script, external URL,
-runtime credential, sibling Track reference or accumulated-composite input.
-
-## Closed rendering vocabulary
-
-The implementation exports `VISUAL_STYLE_NAMES_V1`. A declaration outside that list is
-invalid even when a particular Chromium build happens to understand it. Critical enum-shaped facts
-such as `position`, `display`, `overflow`, `object-fit`, direction and writing mode additionally use
-closed value sets.
-
-The following fail closed:
-
-- unknown CSS properties;
-- `position: fixed` and scrolling layout;
-- `backdrop-filter` or `mix-blend-mode` cross-Track sampling;
-- CSS `url(...)` instead of typed Artifact references;
-- `var(...)`, `env(...)` or `attr(...)` environment-dependent values;
-- declaration-breaking punctuation;
-- animation of properties outside the local animation set.
-
-This is intentionally not "arbitrary CSS in JSON". The CSS-shaped representation is merely the
-serialized instruction vocabulary of the locked browser target.
-
-## Validation ownership
-
-Validation has four separate owners:
-
-1. `@narratage/visual-ir` owns the closed vocabulary and identity;
-2. `@narratage/composition` owns Track schemas and intrinsic self-containment validation;
-3. Film and final-render Producers receive ProgramSpace on explicit graph edges and validate
- relational frame-domain facts without adding lineage fields to Track values;
-4. the selected local or hosted render Endpoint locks the actual renderer implementation and
- validates the produced media fact.
-
-Core contains no video-specific branch. Its generic admission path invokes the installed
-Type-owner validator for intrinsic tree, style and Artifact facts. Relations involving another
-value remain ordinary multi-input Producer checks.
-
-## Extension law
-
-Installing a new visual component must not modify this protocol. The component either:
-
-1. lowers its Program into existing elements and styles; or
-2. materializes the visual it owns as a typed `CompositableSurface` and contributes that Surface.
-
-Only a capability that is both broadly reusable and impossible to represent through those two
-paths may justify changing the `svml.visual-ir@1` contract. While Narratage is pre-release the
-identifier remains `@1`, with exact implementation identity carried by digests and package locks.
-Adding a Text style, Caption mode, Media recipe, transition or Provider never qualifies by itself.
-
-Old-system Text three-box, Caption range/cue/content and Media content-frame sampling models are
-therefore package-owned Programs and lowering evidence. They are expressiveness witnesses for this
-IR, not fields added to the IR.
-
-## Freeze evidence
-
-The identifier and closed vocabulary are executable. Exact Text fonts, multiline decoration,
-Path/Mask/overflow, Media focal sampling and alpha now have real browser evidence; a separately
-installed non-native Text package crosses the typed Surface waist without changing Core, Film or
-HyperFrames. The repository-internal `@1` waist is frozen because all four independent gates now
-execute:
-
-1. the generic Need Receipt binds the locked Endpoint implementation/configuration/Runtime closure,
- while HyperFrames' receipt-covered attestation binds the exact browser bytes locally or the
- explicitly configured immutable renderer deployment digest remotely;
-2. the shared media executor decodes Surface bytes and proves dimensions, frame timing, SDR/sRGB
- compatibility and opaque/straight-alpha facts before local rendering; a renderer Endpoint without
- that verifier must decline a Surface-bearing document;
-3. Deck and Ranking provide the remaining package-owned Track expressiveness witnesses;
-4. the repository-wide boundary audit proves that Composition contains no package-family/provider
- discriminator and the terminal compiler contains no hidden cross-Track sampling path.
-
-This freezes the protocol for subsequent package work in this repository. It is not a statement
-that npm packages have been published, that every author Surface is frozen, or that a future public
-release may skip its distribution and security review.
diff --git a/docs/zh/guide/spec/audio-track.md b/docs/zh/guide/spec/audio-track.md
deleted file mode 100644
index 37b6888f..00000000
--- a/docs/zh/guide/spec/audio-track.md
+++ /dev/null
@@ -1,214 +0,0 @@
----
-title: SVML Audio Track 作者面
-description: 通用官方 Audio Track 包的已实现预发布合同,含自描述作者 Surface 与精确采样域 lowering。
----
-
-# SVML Audio Track 作者面
-
-状态:通用官方 Audio Track 包的已实现预发布合同,包含它的自描述作者 Surface、精确的采样域 lowering,以及本地 / 远端执行见证。它还不是冻结的公开 ABI。
-
-## 目的
-
-`@narratage/audio-track` 让作者把已经选定并归一化的音频素材放入 ProgramSpace。音乐、音效、额外口播和源音频都是普通的同级项。这个包不拥有任何有特权的基础泳道,Composition 通过既有的
-[`AudioTrack`](./track-composition.md) 终端合同来混合它的结果。
-
-这是一个视频领域的作者包,不是 Core 原语、Runtime 服务或 Provider。分工如下:
-
-```text
-explicit media source ─> inspect/select/normalize ─> SynchronizedMedia
- │
-Narrative points / ProgramSpace ─> temporal window ─────┼─> Audio Track Program
-source trim ─> occupancy ─> level/fades ────────────────┘ │
- ▼
- AudioTrack
- │
- ▼
- audio plan -> Provider
-```
-
-Core 看到的是固定的图输入和一个输出 Record。它不知道 clip、bus、音乐、语音、循环、淡变或混音。
-
-## 1. 输入的事实
-
-每个音频条目都通过显式的图边获得素材。第一个实现应当消费来自 `@narratage/media` 的 `SynchronizedMedia`,并要求它归一化后的音频成员:
-
-- 48 kHz;
-- 立体声;
-- 有符号 16 位 PCM WAV;
-- 精确的采样帧数;
-- 保留原始电平。
-
-仅仅含有 AAC 的 MP4 不是音频条目。作者图必须显式地检视容器、选择一条音频流并做归一化。同理,一个生成出来的语音文件不会仅因为里面有人声就变成 `SpeechAudioBasis`。
-
-Audio Track 的 lowerer 可以从归一化后的采样帧数推导终端 Artifact 的时长。它不得把源身份、Narrative 摘要、
-provider 名称或其他血缘信息拷贝进 clip。这些关系仍然由图边和 Derivation 承载。
-
-不要仅仅为了缩短一条边就引入新的公开 `AudioMaterial` Type。如果将来确有多个互不相关的包需要一个一等的规范音频值,那个聚焦的 intrinsic Type 应当放在 `@narratage/media`;它仍然不属于 Core,也不属于某个大杂烩式的 contracts 包。
-
-## 2. 作者 Program
-
-这个包拥有一个任意长度的独立条目集合。在概念上每个条目具有:
-
-```ts
-type AudioItemProgram = {
- readonly id: string;
- readonly source: SynchronizedMediaInput;
- readonly temporal: TemporalBinding;
- readonly trim?: AudioSourceTrim;
- readonly occupancy: AudioOccupancy;
- readonly mix: AudioItemMix;
-};
-```
-
-这些都是包级别的关注点。它们不是 Core Operation 或终端 `AudioTrack` 合同上的新字段。
-
-示意性的作者 Surface 可以保持紧凑:
-
-```xml
-<audio:Track id="mix" space={speech.space}>
- <audio:Clip
- source={music.media}
- during={program}
- playback="loop"
- gain={0.28}
- fade-in="800ms"
- fade-out="1200ms"
- />
-
- <audio:Clip
- source={impact.media}
- at={story.moment.reveal}
- for="420ms"
- playback="once"
- />
-</audio:Track>
-```
-
-这种写法只是示意,不是冻结的 Surface。`during`、`at` 和 `for` 必须 lower 到共享的点表达式与窗口投影代数;它们不能创建第二套时间引擎。
-
-## 3. 时间放置
-
-音频使用与 [`track-authoring.md`](./track-authoring.md) 相同的规则:
-
-1. 定位 Selection、Moment 或 Program 点;
-2. 展开 `one` 或 `each` occurrence;
-3. 投影出一个有方向的候选窗口;
-4. 与 ProgramSpace 求交并校验;
-5. 应用音频的源 trim 与 occupancy;
-6. 把目标放置与源采样一起 lower 到精确的 48 kHz 采样域。
-
-音频条目相互独立。重叠的投影窗口会混合;它们绝不会按书写顺序被裁剪、被自动拼接、被去重,也不会被当成互斥的备选方案。多个 Audio Track 值的行为,与一个 Audio Track 里多个重叠条目完全相同。
-
-ProgramSpace 的帧边界由媒体管线已经拥有的那一条确定性规则转换为 48 kHz 采样边界。作者包不得再实现另一条秒到采样的取整路径。
-
-## 4. 源 trim
-
-源 trim 定义 occupancy 之前的有效固有区间。它是精确归一化采样域内的一个显式起点和可选终点。面向作者的秒或帧只被量化一次到采样边界。
-
-lowerer 拒绝:
-
-- 负值或越出源范围的边界;
-- 不大于起点的终点;
-- 空的有效区间;
-- 不可用的音频成员;
-- 对源时长的第二种包内私有解释。
-
-源 trim 不是目标时间。移动目标窗口绝不会悄悄改变作者声明的是哪一段源区域,除非所选的 occupancy 策略显式地对它做了对齐或重定时。
-
-## 5. 音频 occupancy
-
-音频有它自己诚实的 occupancy 词汇。它不得复用视觉领域那种“把一个采样永远保持住”的虚构。
-
-```ts
-type AudioOccupancy =
- | { readonly mode: "once"; readonly align: "start" | "end" }
- | { readonly mode: "loop"; readonly align: "start" | "end" }
- | {
- readonly mode: "stretch";
- readonly minRate: number;
- readonly maxRate: number;
- readonly pitch: "preserve";
- };
-```
-
-法则如下:
-
-| 策略 | 有效源短于窗口 | 有效源长于窗口 |
-|---|---|---|
-| `once/start` | 在开头播放,随后静音 | 播放源头部并在窗口结束处截断 |
-| `once/end` | 先静音,然后播放到结尾 | 播放源尾部并在窗口结束处结束 |
-| `loop/start` | 从源头部相位开始重复 | 播放源头部并截断 |
-| `loop/end` | 选择相位使源尾部与窗口结束对齐 | 播放源尾部 |
-| `stretch` | 放慢以铺满整个窗口 | 加快以铺满整个窗口 |
-
-`once` 取代了旧的含混的 `native` / `finish` 行为。尾部对齐不是倒放。`loop/end` 改变的是循环相位,它不会反转采样。`stretch` 保持音高,必须停留在作者声明的速率界限内,并且宁可失败也不会悄悄用更快的速率再截断。旧的
-`fit_base` 规则——最多加速 1.1 倍然后把剩下的悄悄切掉——被废弃。
-
-显式的静音由该区间上不存在 clip 来表示。当被填充的静音字节本身有意义时,给文件补静音是一次独立的媒体变换。
-
-## 6. 电平与淡变
-
-第一版作者合同只需要三个正交控制:
-
-- 非负的线性 `gain`;
-- 在可听的已渲染区间上生效的 `fadeIn`;
-- 在可听的已渲染区间上生效的 `fadeOut`。
-
-零增益就是显式的静音。UI 或 recipe 可以暴露分贝,但换算到终端线性增益是确定性的,并且由包拥有。淡变时长必须非负,不能超过可听区间,也不能偷偷扩张目标窗口。
-
-不存在任何隐式的响度归一化、限幅器、压缩器、sidechain 闪避、交叉淡变或自动铺底音乐行为。这些都是合理的未来音频处理或混音 Program,但它们必须消费显式输入,并让自己的行为对作者可见。Runtime 或 Provider 不得把它们当作环境偏好加进来。
-
-原来的终端 `bus: speech | music | sfx | source` 标签已被移除。它对渲染没有任何影响,因此也不承载任何诚实的含义。将来的混音路由合同必须有明确的消费者和图行为;仅仅把一个 clip 命名为 `music` 不能导致闪避。
-
-## 7. Lowering 与执行
-
-确定性的作者包把每个实现出来的条目 lower 成一个普通的终端 `AudioClip`:
-
-- 来自时间投影的精确目标 `{ startSample, endSampleExclusive }`;
-- 来自归一化输入的规范 WAV Artifact;
-- 来自 trim / occupancy 的精确源采样区间、循环相位和播放速率;
-- 来自呈现的保持音高意图、线性增益和精确的采样级淡变。
-
-Composition 会针对显式连接的 ProgramSpace 校验 Audio Track。随后媒体管线把所有同级 Audio Track 编译成一份内容寻址的 `AudioProgramPlan`。本地 FFmpeg、Lambda 或另一个 Provider 执行的是同一份 plan;Provider 的部署位置不能重新解释这次混音。
-
-即使包内拥有很多 clip,它也只发出一个 `AudioTrack`。那是一个作者组件的输出,不是有特权的泳道。天然同时拥有一次视觉命中和一次声音命中的包,可以从一个 Fragment 发出同级的 `VisualTrack` 和 `AudioTrack` 输出,而 Film
-通过普通的边接收两者。
-
-## 8. 旧系统迁移审计
-
-保留:
-
-- 任意音频输入;
-- Program 全长、Selection、Moment 和绝对放置;
-- 重叠的音乐 / 音效 / 人声贡献;
-- 逐条目的增益与淡变;
-- 单次播放、循环,以及有界的保持音高拉伸;
-- 显式的源 trim 和起点 / 终点对齐。
-
-废弃:
-
-- 把动态编号端口当作保存下来的作者事实;
-- 音频上的 `z_index`;
-- 藏在节点执行器里的隐式全长时间;
-- 按配置行优先级做的裁剪;
-- `fit_base` 及其“先加速再静默截断”的行为;
-- 音频的 `fill: freeze`;
-- 自动提取视频里恰好含有的任何音频;
-- 无实际作用的 bus 标签和隐藏的母带处理。
-
-## 9. 实现与验收
-
-共享的 Temporal 包和冻结前的终端 Audio 候选实现现已完成,且没有引入 Core 分支、Provider 家族分支或有特权的
-Film 泳道:
-
-1. **已实现:** 加入 `@narratage/audio-track` 作者包和自描述 Surface;
-2. **已实现:** 消费显式归一化的 `SynchronizedMedia` 输入;
-3. **已实现:** 采样级精确的 trim 和全部 occupancy 法则;
-4. **已实现:** lower 到同级的终端 `AudioTrack`;
-5. **已实现:** 针对任意条目数量以及 Selection / Moment 的 `one` / `each` 的图测试;
-6. **已实现:** 针对更短 / 相等 / 更长的源、两种对齐、循环、拉伸界限和淡变的采样级测试;
-7. **已实现:** 两个重叠条目与两个同级 Audio Track 产出相同的计划混音事实;
-8. **已实现:** 本地 FFmpeg 与远端 Lambda Provider 收到完全相同的内容寻址 `AudioProgramPlan`;真实 FFmpeg
- 见证还会逐采样检查精确的循环相位。
-
-如果加入这个包需要 Core 分支、Film 音频家族、新队列、SVML 中出现 Provider 名称或一条隐藏的 Base 音频规则,那么这次迁移就是不完整的。
diff --git a/docs/zh/guide/spec/author-surface-binding.md b/docs/zh/guide/spec/author-surface-binding.md
deleted file mode 100644
index c60c4fe9..00000000
--- a/docs/zh/guide/spec/author-surface-binding.md
+++ /dev/null
@@ -1,138 +0,0 @@
----
-title: Author Surface Binding
-description: 已实现的编译器合同,覆盖受信任的包自有 Surface;第三方隔离尚未实现。
----
-
-# Author Surface Binding
-
-状态:已实现的编译器合同,覆盖受信任的包自有 Surface;第三方隔离尚未实现。
-
-## 1. 目的
-
-通用的 Markup Frontend 持有元素语法。包自有的 Surface 持有某个被 import 的标签*具体是什么意思*。SVS
-只持有自己的样式表语法,并产出通用的不可变 Recipe record。Markup、SVS 和 Core 都不持有 Caption、Media、
-Film 或其他任何包的属性词汇。
-
-两者之间的桥是编译期引用检查:像 Caption 这样的 Surface 必须能够读取一个被显式引用的公开 Recipe,并在
-Core 图构建之前把它下降为名义的 `CaptionProgram`。
-
-本文定义这座桥,同时不引入全局 Recipe 注册表,也不引入 Runtime 步骤。
-
-## 2. 有序的编译边界
-
-一个 Source Closure 按以下顺序编译:
-
-```text
-recursive imported SourceUnits
- -> admitted public Records and symbolic component exports
- -> current document generic syntax
- -> package-owned Surface compilation
- -> typed authored Records + AuthorComponents + locked GraphFragments
- -> parser-independent Author linking
- -> Core CompiledGraph
-```
-
-因此,被 import 的 `.svs` SourceUnit 一定先于消费它的 `.svml` SourceUnit 完成编译。消费方的 Surface
-拿到的是结构化的元素,以及一个只针对该元素中显式书写的引用的 resolver。
-
-## 3. 引用解析结果
-
-一个已解析的 Surface 引用包含:
-
-```text
-path source spelling used by the element
-ref hygienic AuthorValueRef
-type exact nominal TypeRef
-record? defensively copied TypedRecord when the public export is already an authored Record
-```
-
-组件输出没有编译期 Record,因此省略 `record`。如果某个 Surface 需要检查值本身(例如 Recipe 解码器),它必须要求一份作者 Record,并在引用无法解析、Type 不对或此时还没有值时,于 `check` 阶段失败。
-
-只有被 import 的 SourceUnit 的公开 record export 会被暴露出来。私有的子 record 不属于 resolver 的范围。每一份返回的 Record 都是防御性的规范化副本,因此 Surface 代码无法改写已经编译完成的子 SourceUnit 或此前的声明。
-
-## 4. 包的归属
-
-对于这段源码:
-
-```svml
-<caption-fine:Style recipe={studio.caption.dialogue}/>
-```
-
-各方职责如下:
-
-```text
-@narratage/svs
- parses studio.svs
- produces generic @narratage/svs#Recipe
-
-@narratage/markup
- parses the element and the whole-value reference
- resolves only the public imported binding
-
-@narratage/caption Surface
- requires @narratage/svs#Recipe
- validates Caption-owned properties
- emits nominal @narratage/caption#CaptionTrackProgram
- emits the AuthorComponent and locked GraphFragment
-
-Author compiler
- resolves symbolic component dependencies and types
-
-Core
- sees only typed Records, LogicalOutputs, Candidates and Operations
-```
-
-不存在一张通用属性表。另一个包完全可以通过另一个被显式 import 的 Surface 来解释同一份通用 Recipe,并产出不同的名义 Program。Surface 永远不由 Recipe 的形状来选中;是作者写下带命名空间的组件时选中它的。
-
-## 5. Surface 输出法则
-
-一个 Surface 只能产出:
-
-1. 精确 Type 出现在其锁定 Manifest `outputs` allowlist 中的作者 Record;
-2. Fragment 摘要由该次解码结果贡献的 AuthorComponent;
-3. 不可变的 GraphFragment;
-4. 非语义的 source-map 事实。
-
-Frontend 用源码与 Frontend Closure 身份为作者 Record 的来源封印,校验其结构 Type,并在链接前让它通过
-Host 的准入闸门。Surface 不能注入未声明 Type 的 Record。
-
-由 Recipe 推导出的带类型 Program,应当包含其所属包做语义校验所需的任何源码摘要。它的作者来源已经绑定了
-Surface 实现摘要和递归编译得到的 Source Closure。别名写法与文件系统位置的变化可能改变源码身份,但不会改变带类型的 Program 值,也不会改变最终 Graph 的含义。
-
-## 6. 前向引用法则
-
-两种情况被有意区别对待:
-
-- 图依赖可以保持为符号化的 `component.output` 引用,在整份文档收集完毕之后再解析,因此普通组件的前向
- 引用依然有效。
-- 编译期的值检查要求 Record 已经完成作者编译。被 import 的源码总是可用的;而同一文档内的值定义,必须
- 出现在检查它的 Surface 之前。
-
-这不是某种 parser 模式或 Runtime 模式,而是取决于编译此刻是否需要那个值,还是稍后只需要一条带类型的图边。
-
-## 7. 它不是什么
-
-Surface Binding 不是:
-
-- Core 的 Type 注册表;
-- 隐式转换搜索;
-- Runtime Producer 或付费 Operation;
-- Provider 能力;
-- 全局 SVS 解码器;
-- 读取凭据、文件、网络或 Store 的许可。
-
-当前的 Host 只执行显式注册的受信任 Surface 实现。要让任意社区 Surface 代码可以执行,必须先具备按字节锁定的代码、权限强制和隔离。
-
-## 8. 可执行的证明
-
-与视频无关的 `example.recipe-card` fixture 证明了这条边界:
-
-1. SVS 产出通用 Recipe;
-2. Card Surface 只读取自己显式写下的公开引用;
-3. Card 在作者编译期校验 `fill` 与 `padding`;
-4. Card 产出名义的 `CardAppearance` Record;
-5. 运行时 Producer 消费的是 `CardAppearance`,不是 `SVS Recipe`;
-6. 非法颜色在 BuildPlan 或任何外部操作出现之前就失败;
-7. 只改源码 import 别名时,Program Record 与 Graph 身份保持不变。
-
-视频相关的包必须使用同一套机制,而不是另开一条视频专用的编译器通路。
diff --git a/docs/zh/guide/spec/caption-fine.md b/docs/zh/guide/spec/caption-fine.md
deleted file mode 100644
index db141901..00000000
--- a/docs/zh/guide/spec/caption-fine.md
+++ /dev/null
@@ -1,189 +0,0 @@
----
-title: Fine Caption Style 族
-description: 已实现的预发布 Style 族,所声明的表面均有完整的浏览器证据。
----
-
-# Fine Caption Style 族
-
-状态:已实现的预发布 Style 族,所声明的表面均有完整的浏览器证据。
-
-## 目的与边界
-
-`@narratage/caption-fine` 是官方的、无字段的细粒度 Caption Style 族。这里的 “Fine” 指的是:不可变的显示 Atom 被规划成短的语义 Cue,再经由一套确定性视觉系统渲染。它**不**表示 `important`、语义强调分类或逐词的随机样式。
-
-这个包消费三份已经存在的事实:
-
-1. `CaptionProgram`,它给每个显示 Word 指派一份完整的 Style;
-2. `TimedCaptionProjection`,它给每个 Cue 和每个完整 Atom 一个已证实的时间窗口;
-3. `CaptionDisplaySequence`,它保留作者精确的可见文字与标点。
-
-它输出一条普通的对等 `VisualTrack`。它绝不修改文字、绝不调用模型、绝不推断 Dual Text Atom 内部的时间,也绝不往其他合同里塞 Caption 元数据。
-
-| 归属 | 职责 |
-|---|---|
-| `@narratage/caption` | 通用 Style 外壳、Cue/字段 Plan 法则、全覆盖 Style 赋值与 Atom 时间 join |
-| `@narratage/caption-fine` | Fine Recipe schema、layout、Paint、局部 motion 与 VisualTrack lowering |
-| `@narratage/media` | `FontArtifactRef` 这类精确可复用的字节,绝不涉及 Fine 的排版策略 |
-| `@narratage/visual-ir` | 封闭的、无代码的终端 element/style/keyframe 词汇 |
-| Core | 只负责图的 demand、校验、派生和执行;不含任何 Caption 含义 |
-
-只要现有公开合同还能表达结果,改动 Fine 的阴影、锚点、卡拉 OK 模式或 Cue 盒子,就不得修改 Caption、
-Media、Composition、Visual IR 或 Core。
-
-## 完整的正交表面
-
-Fine 拥有一棵单字体的 Cue/Atom layout 树。它的各个维度互相独立:一份 Recipe 可以同时组合渐变字形、当前词盒子、拖尾文字、下划线和局部 motion,而不必换一个渲染器。只有当某种外观改变了 layout 树、或者需要 planner 字段时——例如双字体的编排式排版——才有理由新开一个包;仅仅增加 Paint 或局部 motion 不算。
-
-### 1. Cue 规划
-
-- `cue-min-words`
-- `cue-max-words`
-
-planner 只能在完整 Atom 之间切分。不可分割的作者 Atom 可以超过偏好的最大值;绝不会为了凑数字而把它拆开。
-
-### 2. 放置与布局
-
-- 归一化的 `x`、`y`、`width`
-- `anchor-x: left | center | right`
-- `anchor-y: top | center | bottom`
-- `align: left | center | right`
-- `direction: ltr | rtl`
-- `line-height`、`letter-spacing`、`word-gap`
-- 绝对的 `stack-order`
-
-Cue 只会在 Atom 之间自然换行。同一个 Atom 内的 Word 绝不会被拆到不同行。Fine 没有 `max-lines` 属性:硬性行数限制要么丢弃作者文字,要么把浏览器测量偷偷带进确定性 lowering。Fine 绝不裁剪、绝不截断。作者通过 Cue 词数边界、宽度和字号来控制密度。如果确实需要一条硬性布局断言,它必须是一个单独连接的组件,其失败在图中是显式的。
-
-### 3. 排版
-
-- Recipe:`size`
-- `text-transform: none | uppercase | lowercase`
-
-text transform 只是终端呈现。它绝不改变 `CaptionDisplaySequence`、planner 输入、Word/Atom 身份、语音对应或时间。
-
-family、weight 和 style 被刻意排除在 Recipe 之外。每个 Style 都必须从 `@narratage/fonts-open` 引入精确的已安装字面,或用 `<media:Font>` 声明自定义/品牌字节。`font=` 接受一个精确字面加上有序的、
-Style 所属的 Fallback 子节点,或者一个可复用的通用 `FontStackRef`。开放字体的紧凑写法是:
-
-```svml
-<fonts:Stack id="caption-fonts" family="inter" weight="700" style="normal" emoji="color">
- <fonts:Fallback family="noto-sans-sc" weight="700" style="normal"/>
-</fonts:Stack>
-<caption-fine:Style id="primary" recipe={studio.caption.primary} font={caption-fonts}/>
-```
-
-字体这条边是 family、weight 和 style 的唯一真相来源。fallback 保留它自己诚实的字面元数据——例如 700
-字重的拉丁主字面可以搭配 400 字重的符号 fallback。完全重复的字面会被拒绝。一个逻辑字面可以包含多个按内容寻址的 Unicode-range 源,已安装的中日韩与 Emoji 字体就是如此。Fine 把 `FontStackRef` 展开成同样有序的精确字面,并且只把它们放在自己的终端文本 element 上;Caption、Core 以及无关的图值保持不变。省略 `font=` 是非法的,因此结果里不会混入任何 Runtime 或机器字体选择策略。
-
-中日韩口播可以直接书写。纯显示的 emoji 仍然需要显式的语音对应,例如 `<🌐 | globe>`;Script 会正确地拒绝为一个孤立符号发明口播 token。同时具有文字与 Emoji 呈现形态的字符,使用作者书写的精确 Unicode
-序列(例如带 VS16 的 `☎️`);没有任何 lowerer 会为了强制上色而改写显示文字。这是时间与文字真相问题,不是字体能力问题。
-
-### 4. 基础字形与激活字形的 Paint
-
-- 纯色 `fill`,或 `gradient-from`、`gradient-to`、`gradient-angle`
-- `opacity`
-- 一层描边:`stroke-color`、`stroke-width`
-- 一层阴影:`shadow-color`、`shadow-opacity`、`shadow-x`、`shadow-y`、`shadow-blur`
-- 一层有界的方向性长阴影:`long-shadow-color`、`long-shadow-opacity`、
- `long-shadow-distance`、`long-shadow-angle`
-- 一层辉光:`glow-color`、`glow-opacity`、`glow-blur`
-- 一层基础下划线:`underline: off | always`、`underline-color`、`underline-thickness`、
- `underline-offset`
-
-每个激活态 Paint 名字都带 `active-` 前缀。例如 `active-gradient-from`、`active-long-shadow-distance`
-和 `active-glow-blur` 描述的是激活字形,不改变基础字形。
-
-基础 Paint 与激活 Paint 形状相同。渐变替换该层的纯色填充;它不替换激活叠加层,也不替换卡拉 OK 过渡。投影、长阴影和辉光是互相独立的作者贡献,会被 lower 成一条确定性的 `text-shadow` 声明。长阴影的展开有实现上的上界,因此一份 Recipe 不可能造出无限膨胀的 CSS 负载。任意多层阴影、纹理、斜面和自由挤出仍在
-Fine 之外。
-
-### 5. Cue 盒子 Paint
-
-- 纯色 `background`
-- `border-color`、`border-width`
-- 水平/垂直 `padding`
-- `radius`
-
-禁止背景采样,因为那会让这条 Track 去窥视另一条 Track 的像素。
-
-### 6. 激活通道、卡拉 OK Paint 与时间
-
-- 激活字形状态:`karaoke: off | current | trail`
-- 激活字形过渡:`karaoke-transition: step | wipe`
-- 激活盒子状态:`active-box: off | current | trail`
-- 激活盒子连续性:`active-box-continuity: isolated | joined`
-- 激活盒子 Paint:`active-box-background`、border、padding 和 radius
-- 激活盒子 motion:`active-box-enter`、`active-box-exit`、`active-box-transition-frames`
-- 激活下划线状态:`active-underline: off | current | trail`,以及颜色、粗细和偏移
-
-激活状态与装饰几何被刻意分开。`current` 只激活其实测窗口包含当前帧的那个 Atom。`trail` 让每个已激活的
-Atom 一直保留到 Cue 结束。`step` 整体切换激活字形层;`wipe` 在该 Atom 的实测窗口内逐步揭示它。RTL 会反转 wipe 方向。字形、盒子和下划线可以各自选择不同的状态策略,因此历史上保留的那种行为——拖尾文字配仅当前的 pill——可以直接表达。
-
-isolated 盒子为每个已激活的 Atom 各画一个盒子。joined 拖尾把有序的已激活前缀画成连续的行内片段:同一渲染行上的 atom 共享一个背景,而每个折行的行各自获得自己的端帽。决定这些片段的是浏览器的行布局,而不是上游元数据。Fine 绝不把实测行盒子传播到图里。
-
-卡拉 OK 刻意是 Atom 粒度的。在普通文字里,一个 Atom 通常就是一个可见 Word。在 `<45% | forty five percent>`
-中,作者书写的可见 `45%` 是一个 Atom,作为一个整体激活。在多词显示 Atom 中,整个作者 Atom 一起激活。下游任何包都不会发明作者与音频证据从未提供过的内部 Word 时间戳。
-
-### 7. 分层的局部 motion
-
-- Cue 进入/退出:`none | fade | pop | scale | spring | bounce | elastic | stamp | tilt |
- zoom-blur | flip-x | flip-y | spin | squash | stretch | slide-left | slide-right | slide-up |
- slide-down | blur-in | wipe-left | wipe-right | wipe-up | wipe-down`
-- Atom 进入与退出:同一套一次性词汇,时长独立
-- Atom 揭示:`all | on-start | typewriter`
-- 激活响应与激活盒子的进入/退出:同一套一次性词汇
-- 连续的局部循环:`none | shake | wobble | glow-pulse | breathe | float | pulse | flicker`,
- 作用于 Cue 或激活 Atom
-
-扁平的 Recipe 名字是 `cue-enter`、`cue-exit` 及其各自独立的 `*-frames`;`atom-enter`、
-`atom-enter-frames`、`atom-exit`、`atom-exit-frames`、`atom-reveal`;`active-response`、
-`active-response-frames`、`active-scale`;
-`slide-distance`;以及 `loop`、`loop-target`、`loop-period-frames`、`loop-intensity`。
-
-Cue、Atom 生命周期、激活响应和循环各自拥有独立的包裹层,因此它们的 transform 是叠加组合而不是互相覆盖。typewriter 在一个完整 Atom 内逐个揭示作者书写的显示字素;它是视觉插值,不代表字符级时间戳,也绝不改变 Atom 边界。连续 motion 是有界的确定性帧函数,不含任何随机输入。任意路径运动和随机变化仍在
-Fine 之外;它们属于通用 Typography 或独立的特效包,而不属于 Caption 的预设词汇。
-
-## 一个渲染器,一棵 layout 树
-
-静态输出与卡拉 OK 输出共享这套结构:
-
-```text
-placement
-└── cue motion
- └── cue box
- ├── joined/isolated decoration underlay
- └── cue loop
- └── atom entry
- └── active response
- ├── base words
- └── active glyph/underline overlay
-```
-
-装饰底层只重复那份为了获得诚实的浏览器行片段所必需的、同样的透明字形几何。它不会制造出第二份文字真相。激活叠加层复用同一份 Atom 几何。每个动画都会在 HyperFrames 看到它之前,lower 成有限的、按帧寻址的
-`VisualTrack` keyframe。
-
-## 默认值与兼容性
-
-最初那十五个基线属性仍然是必填的,因此最小的 Recipe 也会对规划、几何、排版和可见盒子给出显式说明。新增维度是可选的,并解析成一个完整的不可变参数对象:
-
-- 左上锚点、LTR、normal 字体、零字距,以及四分之一 em 的词间距;
-- 无描边、无阴影、无长阴影、无辉光、无下划线、无边框;
-- 纯色填充,text transform 为 none;
-- 激活字形、激活盒子和激活下划线均关闭;
-- Cue/Atom/盒子的 motion 与循环均关闭,所有 Atom 可见;启用激活响应通道时,其 scale 默认为 1.08;
-- 启用卡拉 OK 但未指定其他激活 Paint 时,激活填充默认为 `#FFD54A`,其余激活 Paint 维度继承基础 Paint。
-
-未知属性一律拒绝。默认值属于包的实现策略,因此被其实现摘要覆盖;它们不是隐藏的 Runtime 行为。
-
-## 已完成的证据门
-
-完整设计是渐进交付的,而不是拆成互不兼容的版本:
-
-1. **已解析模型与静态 lowering** —— 完整参数、锚点、布局、基础/激活 Paint、Cue 盒子与严格校验。
-2. **定时 lowering** —— current/trail、step/wipe、Cue 淡入淡出、Atom 揭示和激活 scale,全部只使用已
- 证实的整 Atom 时间。
-3. **可复现性证据** —— 有序的精确 Font Artifact 栈、中日韩、emoji/符号 fallback、多行折行、描边、阴影、
- 辉光以及全部四种字形卡拉 OK 模式,都有真实的浏览器/像素见证。`max-lines` 是被刻意拒绝,而不是被
- 推迟。
-4. **表现力 Paint 与 motion 证据** —— 仅当前盒子与拖尾盒子、isolated 与 joined 几何、渐变、下划线、
- 长阴影、text transform、分层的一次性 motion、typewriter 揭示与确定性循环,各自都有终端 IR 见证;
- joined 几何另外还有真实折行的浏览器/像素见证。
-
-没有任何一道门改动 Core 或公共 Caption。精确的字体选择是 Media Font Surface 与 Fine Style Surface 之间一条显式的作者图引用,而不是沿着 Caption 流水线传播的元数据。
diff --git a/docs/zh/guide/spec/caption-program.md b/docs/zh/guide/spec/caption-program.md
deleted file mode 100644
index 235a676a..00000000
--- a/docs/zh/guide/spec/caption-program.md
+++ /dev/null
@@ -1,149 +0,0 @@
----
-title: Caption Program
-description: 预发布的可执行合同。
----
-
-# Caption Program
-
-状态:预发布的可执行合同。
-
-## 1. 范围
-
-Caption 是一个普通的视频组件族。它把不可变的作者显示文字变成一条自包含的 `VisualTrack`。它对
-Script 的措辞、发音、语音对齐、Provider 选择、Runtime 执行和 Composition 顺序都没有任何话语权。
-
-它的两条独立图分支是:
-
-```text
-CaptionDisplaySequence + CaptionProgram ── planner ─────────── CaptionPlan
-speech audio ────────────────────────────── evidence/locator ── CompleteSemanticMap
-
-Display + Correspondence + Plan + Program + Map ── timing ─── TimedCaptionProjection
-TimedCaptionProjection + Display + Program ─────── renderer ── VisualTrack
-```
-
-planner 永远拿不到时间戳。对齐永远不做 Cue 或 Style 判断。两条分支只在显式的 timing 组件处汇合。
-
-## 2. 显示真相:Atom 与 Word
-
-Script 输出一份 `CaptionDisplaySequence`,它带两个有序视图:
-
-- `Atom` 是 planner 可以整体放进一个 Cue 的最小单位,也是能够承载已证实语音时间的最小显示单位;
-- `Word` 是可选的 Style 所属字段的最小作用目标。
-
-每个 Word 恰好属于一个 Atom。Atom 对 Word 的划分恰好覆盖一次并保持顺序。文字与标点始终是作者真相:
-`45%`、`back-and-forth`、`damn!`、`300,000`、`don't` 和 `U.S.A.` 原样穿过这条边。官方 Script reader
-以英文优先,把作者书写的空白当作正常的显示词边界;其他作者包可以发布另一套策略,只要产出同样的公开合同。
-
-对普通文字,一个显示 Word 通常就是一个 Atom。显式的 Dual Text 永远是一个完整 Atom,即使两侧恰好含有相同的词:
-
-```svml
-<New York City | new york city>
-```
-
-该显示序列包含一个 Atom、三个 Word:`New`、`York`、`City`。它并不声称其中任何一个显示 Word 对应任何一个口播 token。隐藏的 Dual Text(`< | um>`)不产生任何显示 Atom。
-
-这是刻意的信息纪律:Script 只使用作者显式给出的对应关系,绝不发明作者没有给出的对应关系。
-
-## 3. 语音对应是一条独立的边
-
-`CaptionCorrespondence` 把每个显示 Atom 映射到一个或多个作者书写的口播 token 身份。它与
-`CaptionDisplaySequence` 相互独立;显示侧的消费者不必携带自己用不上的语音血缘。
-
-对普通文字,每个 Atom 映射到该显示表面在结构上拥有的那些口播 token。对 Dual Text,整个显示 Atom 映射到整段右侧口播区间。不存在公开的“显示 Word 到口播 Word”的映射,也不存在猜出来的内部时间。
-
-Script 还会为显式的 Selection 输出 `CaptionDisplayWordSubset` 值。Role 选择器是同一份显示序列之上的作者面语法糖。子集是有序的,可以选中零个或多个完整 Atom;只选中一个多词 Atom 的一部分属于作者错误。这些值不是时间性的 `NarrativeSelection` 值,也不会引入第二套 `when` 语言。
-
-## 4. 完整的 Style 意图
-
-一个 `CaptionStyle` 是一份完整的作者声明,包含:
-
-- 通用的规划要求:偏好的 Cue 词数上下界,以及零个或多个带类型的 per-Word 字段声明;
-- 一个渲染族身份,以及该族经 Recipe 解析后的完整参数。
-
-公共 Caption 包只拥有 Cue 与字段的数据形状,不拥有字体、盒子、颜色、动画,也不定义字段 id 的含义。一个 Style 族可以不声明任何字段。布尔、枚举和有界数值字段对其他族同样可用;一个 Word 可以携带零个、一个或多个互相独立的字段。
-
-SVS 存放具名的 Recipe 数据。Style 包解释这份 Recipe 并输出一份完整的 Style。应用另一个 Style 会整体替换原 Style;规划与渲染的零散片段绝不会被隐式合并。
-
-## 5. 全覆盖赋值与有序替换
-
-每个 `CaptionProgram` 都有一个显式的 default Style,覆盖完整的显示全集。作者无需伪造 `@whole`
-Selection,也不必写补集。
-
-```xml
-<caption:Program id="captions" display={story.caption} default={plain}>
- <caption:Use role="ALICE" style={alice}/>
- <caption:Use words={story.caption.selection.special} style={impact}/>
- <caption:Mute words={story.caption.selection.private}/>
-</caption:Program>
-```
-
-Use 按源码顺序生效,最后一个命中的整体 Style 替换获胜。Program 存的是显示序列身份和有序的 Word id
-区段,而不是复制出来的文字。一个区段绝不跨越 Segment 或 Turn 边界。所有区段互不相交,对显示 Word 恰好划分一次,且任何区段都不得把一个 Atom 拆开。
-
-`Mute` 消费的是与 `Use` 相同的、Caption 专有的整 Atom 词子集。多个 `Mute` 子节点构成一个有序并集,因此一个 Program 可以隐藏任意多段互不相连的作者区间。`Mute` 是规划之后的可见性意图:被 mute 的 Atom
-仍然留在不可变的显示全集和 `CaptionPlan` 中,绝不会作为指令进入 planner 提示词,也绝不触发 Cue 重新规划。公共的 timing join 会把被 mute 的 Atom 从已经规划好的定时 Cue 中移除,同时保留该 Cue 的身份和原有时间跨度;一个没有任何可见 Atom 的 Cue 不出现在定时视觉投影中。因此各 Style 族的渲染器共享这一行为,不必各自发明自己的遮罩语义。公共包还导出同一个幂等操作,供直接调用渲染器时做防御性处理。
-
-这刻意不是一个通用的时间遮罩。Script 已经把作者的 Selection 投影成了精确的显示词子集,因此 Caption
-无需把词换算成时间、再从重叠窗口里把词猜回来。只选中多词 Dual Text Atom 一部分的 `Mute`,会被与
-Style 应用相同的不可分割规则拒绝。
-
-`display={story.caption}` 是一条真实的图边。planner、timing 和渲染只在自己需要时,各自声明指向同一个值的边。
-
-## 6. planner 合同
-
-planner 接收已解析的区段。对每个区段,它只能:
-
-1. 把连续的完整 Atom 编组成有序的 Cue;
-2. 给每个 Cue 内的 Word 赋予已声明的字段。
-
-Cue 必须对每个区段恰好划分一次。偏好的 Cue 最小/最大词数只是对 planner 的引导,不能让作者文字失效;单个多词 Atom 可以超过最大值,并且必须保持完整。响应中不包含任何替换文字、发音、时间戳或 Style。
-
-Gemini planner 只发送一次可读形式:
-
-```json
-{
- "atoms": [
- ["new"],
- ["york"],
- ["city", "is", "beautiful"]
- ]
-}
-```
-
-外层数组是 Cue 的可切分全集。每个内层数组是一个不可分割的 Atom。其中的字符串是不可变的字段作用目标。稳定 id 留在包内部。响应通过 `atom_count` 消费 Atom;字段在其 Cue 内使用从 1 开始的 `atom_number` 和
-`word_number`。
-
-## 7. 时间投影
-
-`CompleteSemanticMap` 仍然是唯一的全局语音时间真相。timing 组件通过 `CaptionCorrespondence` 解析出每个
-Atom 的显式口播 token 区间,并给整个 Atom 一个实测窗口。一个 Cue 从它的第一个 Atom 开始,到它的最后一个
-Atom 结束。
-
-因此 `TimedCaptionProjection` 只包含定时的 Cue 和定时的 Atom。它不包含 per-display-Word 的时间戳,没有 `estimated` 质量标签,也没有按字符比例的切分。对于多词的 Dual Text Atom,下游渲染器可以自行决定如何排布或动画其内部 Word,但不得声称拥有编造出来的词级时间。任何内容都不会写回 SemanticMap。
-
-## 8. 渲染与合成
-
-每个定时 Cue 选出其解析后 Style 的完整渲染负载。一个 Style 族渲染器返回一条普通的 `VisualTrack`。在首版合同中,一个 Program 内的所有 Style 使用同一个渲染族。
-
-`@narratage/caption-fine` 是第一个官方族。它不声明任何字段,拥有一套正交的单字体 layout/Paint/motion
-系统,包括互相独立的整 Atom 字形、下划线和 Pill 激活。其他包可以定义不同的 layout 树或依赖字段的
-Paint,而无需改动 Script、Caption、Composition 或 Core。
-
-Film 不对 Caption 做特殊识别。它把产出的对等 Track 与 Speech、B-roll、Text 以及第三方 Track 并排折叠。
-
-## 9. 显式的非目标
-
-Caption v1 不定义:
-
-- Script 纠错或替代措辞;
-- 推断出来的 Dual Text 词级对应;
-- 说话人分离或说话人身份;
-- 隐式的模型或 Provider 路由;
-- 自动的缓存复用;
-- 估算出来的显示词时间;
-- planner 可见或由 Core 拥有的 Caption mute 语义;
-- Composition 中的全局 Caption 图层;
-- Core 中的字段专有语义。
-
-Target 与 Candidate 仍然是外部的运行意图。它们不改变 Caption 数据,也不改变 planner 语义。
diff --git a/docs/zh/guide/spec/core-kernel.md b/docs/zh/guide/spec/core-kernel.md
deleted file mode 100644
index 5259954c..00000000
--- a/docs/zh/guide/spec/core-kernel.md
+++ /dev/null
@@ -1,71 +0,0 @@
----
-title: Narratage Core Kernel
-description: 当前可执行合同:持久化 Graph、BuildRequest、Plan 与 BuildState 的 wire 格式。
----
-
-# Narratage Core Kernel
-
-状态:当前可执行合同。持久化的 Graph、BuildRequest、Plan 与 BuildState wire 格式都是 `@1`。
-
-Core 是与领域无关的图 Demand 编译器,以及一台经过校验的 Build 状态机。它不解析源码、不加载包、不执行实现、不选择 Endpoint、不访问凭据、不持久化 Artifact,也不认识任何视频词汇。
-
-## 输入
-
-Core 只接受不可变的、已经完成链接的数据:
-
-- 一份已解析的 module closure 与带类型的作者 Record;
-- 一份完整的 `svml.graph@1`,其中包含来自 Author Graph 与 Run Graph 的 Logical Output、彼此独立的
- Candidate 和原子 Operation;
-- 一份包含 Target 的 `svml.build-request@1`。
-
-图的构造、源码解析、包解析、Fragment 展开和 Run Graph 的编写都发生在 Core 之外。Core 在推导 plan
-之前先校验这些数据。
-
-## 图中的值
-
-1. 一个 `LogicalOutput` 是稳定的、作者可见的承诺:身份、名义 Type 和 Primary Candidate。
-2. 一个 `Candidate` 是一份独立的带类型供给。它的根要么是一个不可变的 Provided Value,要么是一个原子
- Operation 的结果。它不隶属于任何 output。
-3. Run 编译器逐条应用显式的 `Satisfaction`:在不可变的 realized graph 中,把对应 Candidate 设为该
- output 的 Primary Candidate。Core 不接受第二个选择来源。
-4. 一个 `Target` 指名本次 Build 所需要的一个 output。
-5. 一个 `Operation` 具有稳定的实例身份、已声明的带类型输入,以及恰好一个原子结果。多结果组件先产出
- 一个 Product,再接确定性的 Projection Operation。
-
-## 编译法则
-
-1. Core 从所有 Target 出发,沿被选中的 Candidate 和 Operation 输入反向推导 Demand。
-2. 选择结果已经冻结在图中;只有可达性决定要做哪些工作。
-3. 被多条边引用的同一个 Operation 实例只纳入一次。不同实例之间绝不做内容去重,即使实现、参数或输入
- Record 完全一致。
-4. Provided-Value Candidate 没有输入边,遍历自然终止。Core 没有 Pin、预览、缓存或历史结果分支。
-5. 一个 Candidate 必须产出使用它的那个 Logical Output 所要求的、完全一致的名义 Type。
-6. 环、悬空引用、未声明端口、重复身份和不兼容 Type 都在执行之前失败。
-7. 产出的 `svml.plan@1` 是有限的、确定的,并按内容绑定到已校验的 Graph 和 BuildRequest。
-
-## 执行法则
-
-1. Core 从冻结的 BuildPlan 发出 Command。Driver 与 Runtime 无法改变 Candidate 选择、依赖拓扑或 Target
- 的含义。
-2. Record、Need、Command、Event、Receipt 和 Derivation 都是不可变的、按内容绑定的事实。
-3. 一份 Derivation 绑定实现身份、输入与输出摘要、Need 请求摘要,以及被接受的 Event 摘要。
-4. 一份外部 Receipt 绑定确切的 Need、请求摘要、履约实现、输出摘要和被接受的 Event。
-5. 序列化的未完成 Command 不可信。Resume 会校验已接受的状态,丢弃旧 Command,并确定性地重新生成下一批
- Command。
-6. 只有当每一个 Target Record 都存在时,Build 才算完成。
-
-## 类型法则
-
-1. Type 由 module 按名义持有,不在某个中心化的 Core union 里注册。
-2. Core 对每一条被准入的 Record 校验锁定的结构 Schema。
-3. Type 的持有者可以锁定一个语义 validator 摘要。一旦存在,该 Type 的每一条 Record 都需要一份 receipt,
- 绑定到确切的 Type、值摘要和 validator 实现。
-4. Graph 持有依赖关系,Derivation 持有确定性血缘,Receipt 持有外部履约来源。
-
-## 边界
-
-`@narratage/protocol` 持有 wire 数据与身份。`@narratage/core` 持有校验、Demand 编译和状态转移。编译器包持有源码与图的构造。Runtime 持有调度、持久化与放置。Endpoint 包履约确切的 Capability。领域包持有共享词汇与 validator。
-
-文件后缀、包名和 TypeScript class 都不授予 Kernel 特权。在 Core 之后发布的领域包,可以只通过 manifest、名义 Type 和普通 Operation 参与进来,不需要重新发布 Core。
-
-Kernel 没有特权终端输出,没有 Film 根节点,没有视频流水线方向,也没有单 Target 限制。
diff --git a/docs/zh/guide/spec/deck-track.md b/docs/zh/guide/spec/deck-track.md
deleted file mode 100644
index cb423b3c..00000000
--- a/docs/zh/guide/spec/deck-track.md
+++ /dev/null
@@ -1,253 +0,0 @@
----
-title: SVML 深度栈 Deck Track
-description: 官方深度栈 Deck 迁移的已实现可执行权威,记录旧 Deck 实现的行为。
----
-
-# SVML 深度栈 Deck Track
-
-状态:官方深度栈 Deck 迁移的已实现可执行权威。它记录了旧 Deck 实现所使用的行为,但它不是冻结的公开 ABI,也不定义通用的 Deck 协议。
-
-## 1. 结论
-
-Deck 不是 Media Item 的一种模式,也不是 Media Track 的第三个原语。它是更高阶的视觉组件,只是它的成员恰好使用了媒体素材。
-
-第一个官方实现是 `@narratage/deck-track`。它诚实的作者含义是一个**深度栈 Deck**:
-
-```text
-ordered Cards
-semantic activation points
-current Card
-visible previous/future neighborhood
-relative-depth layout
-whole-collection state transition
-ordinary VisualTrack output
-```
-
-它与 Media Track、Text、Caption 和 Ranking 是同级关系:
-
-```text
-Media Item / Sequence ─┐
-Depth-Stack Deck ──────┤
-Text ──────────────────┼─> ordinary VisualTrack values ─> Composition
-Caption ───────────────┤
-Ranking ───────────────┘
-```
-
-Core、Film、Composition 和 HyperFrames 都不知道某个贡献来自 Deck。
-
-## 2. 为什么 Deck 在 Media Track 之外
-
-Media Item 回答的是单个素材如何呈现。Media Sequence 回答的是成员如何在同一个共享表面上相互替换。深度栈 Deck 回答的是另一个问题:当当前成员改变时,整个有序集合如何改变姿态。
-
-在同一帧上,一个 Deck 可能同时显示:
-
-- 当前 Card;
-- 若干保留下来的历史 Card;
-- 若干未来的预览 Card;
-- 按相对深度区分的不同偏移、缩放、旋转、不透明度和色调;
-- 一次让每个保留 Card 从旧姿态移动到新姿态的转场。
-
-这是集合状态,而不是源采样、适配或“出场 / 入场”成对交接。把它留在 Media Track 里,会让一个历史遗留的呈现模型看起来像通用的媒体原语。
-
-## 3. 目前还没有通用的 Deck 合同
-
-“Deck”当前是一个设计家族,不是共享的公开 Type。其他有价值的模型可能具有本质不同的拓扑:
-
-- 水平或环形 Carousel;
-- 扇形布局;
-- Cover Flow;
-- 带有一个特殊当前成员的网格;
-- 自由拼贴;
-- 时间轴历史;
-- 看板与条目一同演化的 Ranking 板。
-
-这些包可以定义不同的作者 Surface、校验与 lowering,它们只需要发出终端 `VisualTrack` 合同即可。只有当多个真实实现证明存在一条更小的共同法则之后,我们才应该抽出共享的 Deck 库。仅仅共用“Deck”这个名词,不构成设立中心协议的证据。
-
-因此第一个包应该暴露一个诚实的组件名,例如 `DepthStack`,而不是声称它的字段描述了所有可能的 Deck。
-
-## 4. 包边界
-
-预期的关系是:
-
-```text
-@narratage/temporal point location and exact trigger schedule
-@narratage/spatial Canvas, Placement Frame and fitted geometry
-@narratage/media normalized still/timed/Surface material
-@narratage/media-track focused reusable media-layer lowering implementation
-@narratage/deck-track DepthStack author Program, state and reflow lowering
-@narratage/composition terminal VisualTrack waist
-```
-
-`@narratage/deck-track` 可以复用 Media Track 中聚焦的实现函数,或者后续某个纯实现层的媒体 lowering 库。这种代码复用并不会让 Deck 变成 Media Track 组件,也不会通过 Core 暴露通用的 `MediaProgram` 或
-`DeckProgram`。
-
-这个包不做以下事情:
-
-- 生成 Card 的图像或视频;
-- 选择 Provider 或凭据;
-- 从 URL、目录或媒体元数据推断 Card;
-- 检视另一个 Track 或已累积的下层合成结果;
-- 预留全局的 stacking 区段;
-- 向 Core、Film、Composition 或 HyperFrames 添加 Deck 语义。
-
-每个 Card 的来源、标签、触发点和声音依赖都是显式的图引用。
-
-## 5. 示意性的作者 Surface
-
-语法尚未冻结,但包的归属应当可见:
-
-```xml
-<deck:DepthStack
- id="proof-stack"
- map={timing.map}
- space={speech.space}
- canvas={vertical}
- frame={layout.lower-proof-stack}
- until={story.selection.proof.end}
- appearance={studio.deck.proof}
->
- <deck:Card id="proof-1" source={proof1.image} at={story.moment.proof1}/>
- <deck:Card id="proof-2" source={proof2.image} at={story.moment.proof2}/>
- <deck:Card id="proof-3" source={proof3.video} at={story.moment.proof3}/>
-</deck:DepthStack>
-```
-
-组件本身产出一个普通的 `VisualTrack`。它不嵌套在 `<media:Track>` 里,也不通过 `mode="deck"` 选择。
-
-SVML 拥有拓扑:有序的 Card 成员关系、source 边、trigger 引用、终止点、Frame 以及可选的标签输入。SVS
-拥有可复用参数:可见邻域、深度姿态、frame Paint 和 reflow 运动。SVS 不能创建 Card、隐藏 source 边或选择 Provider。
-
-## 6. 精确的状态模型
-
-设作者声明的 Card 激活帧为:
-
-```text
-p1 < p2 < ... < pN < T
-```
-
-其中 `T` 是显式的终止点。在帧 `f` 上:
-
-```text
-current(f) = greatest i such that pi <= f
-```
-
-在 `p1` 之前,除非有显式的初始状态策略另行规定,组件处于非激活状态。在 `T` 及其之后,除了作者显式声明的退场运动之外,组件同样处于非激活状态。
-
-包会校验作者声明的出现顺序。它绝不会把相等、逆序或缺失的 trigger 点整理成一份看似合法的呈现。
-
-可见性是显式的包参数:
-
-```ts
-type DepthStackVisibility = {
- readonly previous: number;
- readonly next: number;
- readonly wrap: boolean;
-};
-```
-
-每个可见 Card 都会围绕 `current(f)` 得到一个整数相对深度。一个纯函数把该深度映射为它的姿态:
-
-```ts
-type DepthStackPose = {
- readonly x: number;
- readonly y: number;
- readonly scale: number;
- readonly rotationDeg: number;
- readonly opacity: number;
- readonly stacking: number;
- readonly tone?: DeckCardTone;
-};
-```
-
-公开的作者模型应当暴露具名的 pose / visibility Recipe,而不是要求作者把整张表内联重复一遍。
-
-## 7. Reflow 是集合级的状态转换
-
-当 Card `i` 成为当前项时,在旧状态或新状态中可见的每一个 Card 都会得到一个旧姿态和一个新姿态。reflow
-转场以一个精确时长和一条缓动法则在这两个姿态之间插值。
-
-这与 Media Sequence 的交接不同:
-
-```text
-Sequence transition coordinates outgoing member + incoming member
-Deck reflow coordinates the union of old-visible + new-visible Cards
-```
-
-它们不得共用同一个 `transitionStyle` 联合类型。Card 还可以拥有自己的当前项强调效果,但它不能改动属于另一个 Track 的像素。
-
-渲染器把状态和姿态求值为绝对帧号的纯函数,不保留可变的“上一帧”状态。因此任意帧渲染、分块 Lambda 渲染和乱序帧求值都保持确定性。
-
-## 8. Card 素材与播放
-
-每个 Card 都可以复用与 Media Item 相同的显式 layer、Content Fit、裁剪和 frame Paint 法则。这属于实现复用,不是作者模型的继承。
-
-带时间的 Card 源必须定义它在非当前项时如何采样:
-
-```text
-future preview hold source head, or sample an explicitly selected visible clock
-current Card sample according to its authored active-phase occupancy
-past trail hold source tail, continue, or hide according to explicit policy
-```
-
-静态素材不需要播放策略。视频绝不能在仅作为未来预览时悄悄播完。
-
-Card 标签是可选的显式 Text 输入。Deck 包可以复用 typography lowering,但它不能从文件名、URL 或隐藏的图像元数据推导 Text 事实。
-
-初版 DepthStack 只有视觉。如果后续某个 Deck 组件要暴露源音频,它必须以独立且显式的 `AudioTrack` 投影提供,并带有作者声明的当前成员 / 交接法则;音频绝不能仅仅因为某个 Card 的容器里含有音频流就变得可闻。
-
-## 9. Stacking
-
-Deck Track 没有 Track 级 z-index,也不创建 stacking context。当同级 Track 需要插入其间时,看板 / 背景与各个 Card 可以 lower 成各自独立的绝对 stack Present。
-
-相对深度只决定自有 Card 之间的顺序。reflow 期间的临时顺序也只存在于这一显式关系内部,它不会分配全局 z 段。
-
-## 10. 从旧实现迁移
-
-从旧 Deck 实现中保留:
-
-- 作者显式声明的 Card 顺序;
-- 由语义 trigger 驱动的当前状态;
-- 可配置的 previous / future 可见性;
-- 有限或显式的环绕行为;
-- 深度偏移、缩放、交替旋转、不透明度和色调;
-- 整组的入场、持续与退场运动;
-- 全组插值 reflow;
-- frame Paint、内边距、圆角、描边和阴影;
-- 确定性的帧内局部求值。
-
-废弃:
-
-- 把 Deck 当作 B-roll 或 Media 的一种模式;
-- 隐式的 `linkNext` 拓扑;
-- 藏在“一般位于人脸下方”背后的放置语义;
-- 从源元数据自动生成标签;
-- 下层合成结果参与的转场;
-- 泳道级的全局声音默认值;
-- 在 Core、Film、Composition 或 HyperFrames 中的任何注册表条目。
-
-## 11. 实现与验收顺序
-
-1. 实现共享的 Temporal 与 Spatial 切片;
-2. 暴露 `DepthStack` 和显式的有序 Card;
-3. 求解精确的 trigger 状态与有限可见性;
-4. 为 Card 内容复用聚焦的媒体层 lowering;
-5. 加入纯粹的相对深度姿态求解;
-6. 加入集合 reflow 与整组生命周期运动;
-7. 证明静态与带时间素材的非激活播放行为;
-8. 迁移一个真实的旧系统 Deck,并且只有在验收矩阵通过之后才冻结。
-
-第 1–8 步在 `@narratage/deck-track` 中执行。包测试覆盖下文列出的每一种状态、校验、播放、标签和扩展场景。真实的 HyperFrames 见证会渲染一个旧风格的三 Card 深度栈两次——一次单 worker、一次三个分区 worker——并逐字节比较每一帧解码结果,同时证明每个作者声明的当前状态都被绘制出来。这完成了 Deck 包的验收矩阵,独立的通用 E1–E8 终端 VisualTrack / Visual IR 关卡现在也已通过。
-
-要求的证据包括:
-
-- 只有 previous、只有 next 以及混合的邻域;
-- 零个、一个和多个可见邻居;
-- 有限边界与显式环绕;
-- 缺失、相等和逆序 trigger 的拒绝;
-- 深度姿态、交替旋转、不透明度和色调;
-- 硬切与插值两种全集合 reflow;
-- 静态与带时间 Card 的采样法则;
-- 可选的精确字体标签;
-- 整组入场 / 持续 / 退场且不覆盖 Card 姿态;
-- 顺序、分区和乱序渲染下逐字节一致的帧;
-- 安装另一个 Deck 包时,无需改动 Core 或第一个 DepthStack 包。
diff --git a/docs/zh/guide/spec/media-track.md b/docs/zh/guide/spec/media-track.md
deleted file mode 100644
index 059a3c5a..00000000
--- a/docs/zh/guide/spec/media-track.md
+++ /dev/null
@@ -1,632 +0,0 @@
----
-title: SVML Media Track 创作
-description: 统一官方 Media Track 的可执行预发布权威文档。Item/Sequence。
----
-
-# SVML Media Track 创作
-
-状态:统一官方 Media Track 的可执行预发布权威文档。Item/Sequence 两个 Surface、定时素材、有序 Layer、运动、Handoff、独立的音频投影以及受限的 Speech 投影均已实现。原先的
-`@narratage/broll` 垂直切片已被移除,而不是作为兼容层保留。本合同尚不是冻结的公开 ABI。
-
-## 1. Conclusion
-
-全画幅插叙、下方证据卡、角落图片、透明贴纸、动图、短视频以及普通的替换序列,并不是各不相同的终端 Track family。它们只是同一组媒体关注点的不同组合:
-
-```text
-source material
-target time / trigger schedule
-source-time occupancy
-Placement Frame
-ordered local layers and Content Fits
-clip and frame Paint
-entry / sustain / exit motion
-per-layer sampling motion
-absolute stacking
-optional owned pairwise replacement relationship
-optional explicit audio projection
-```
-
-它们属于同一个视频领域包 `@narratage/media-track`,并且全部下降为普通的对等 `VisualTrack`;只有在作者显式声明时,才会同时下降为 `AudioTrack` 值。
-
-一个包不等于一个带条件分支的超级组件。作者 Surface 暴露两个语义不同的组件:
-
-```text
-Media Item one independent presentation with one projected window
-Media Sequence ordered members replace one another on one shared surface
-```
-
-`Item` 和 `Sequence` 共享聚焦的值族和下降代码。它们之所以是两个独立组件,是因为拓扑、校验和端口都不同。不存在 `mode="broll | sequence | overlay"` 这样的字段,也不存在“选了某个 mode 就冒出另一个隐藏组件”的步骤。
-
-深度堆叠的 Deck 被刻意排除在本包之外。它是更高阶的集合状态组件,而不是媒体原语。官方的迁移模型在
-[`deck-track.md`](./deck-track.md) 中独立规定;未来的 Carousel、Fan、Grid 或其他 Deck 包并不必须采用那套模型。
-
-“B-roll” 仍是有用的剪辑术语或 Recipe 名称,但它不是公开的数据 Type。
-
-## 2. Boundary
-
-统一后的包是一个作者包,同时也是确定性编译器:
-
-```text
-@narratage/media intrinsic media facts and normalized sources
-@narratage/media-pipeline inspection, stream selection and normalization vocabulary
-@narratage/temporal points, occurrences, windows and trigger schedules
-@narratage/spatial Canvas, Placement Frame, Content Fit and fitted geometry
-@narratage/media-track Item / Sequence author Programs and lowering
-@narratage/deck-track independent depth-stack Deck authoring and lowering
-@narratage/composition terminal VisualTrack / AudioTrack waist
-```
-
-它不做这些事:
-
-- 生成图像或视频;
-- 挑选模型或 Provider;
-- 检查凭据、URL 或本地可执行文件;
-- 让 Runtime 或 Provider 自行发明流选择策略;
-- 拥有 Program 时钟或 Canvas;
-- 读取兄弟 Track 或已累积的下层合成结果;
-- 往 Core、Film、Composition 或 HyperFrames 里添加 B-roll、Sequence、Deck 或 Speech 分支。
-
-Core 只看到固定的图 Operation、Record 和边。Runtime 执行任何显式的媒体检查/归一化 Need,或
-Surface 物化 Need。Provider 不会重新解释 placement、motion、transition 或音频策略。
-
-## 3. Source truth
-
-每一个可见 sample 都通过显式的图边进入。作者 Surface 区分四种形态,而不是从单一 Blob 类型去猜:
-
-- 静态图像素材,带内容寻址的 Artifact 和精确的显示方向尺寸;
-- 生成的/原始视频素材,使用 Track 的 ProgramSpace 帧率,通过通用 Media Pipeline 下降,并带显式的
- 纯视觉或 A/V 策略;
-- 已归一化的定时视觉素材,带精确帧率、帧数和尺寸;
-- 带类型的 `CompositableSurface` 素材,带精确的静态/逐帧时序和 alpha 模式。
-
-固有事实的所有者仍然是 `@narratage/media`。Media Track 包内部可以使用包内的已解析联合类型,但不得再造一个通用媒体 Type,也不得把 provider、源路径、Narrative 摘要或上游血缘复制进每一个 item。
-
-按文件名推断是非法的。具体来说:
-
-- 动图 GIF 或动图 WebP 不能继续作为由浏览器计时的 `<img>` 存在,因为并行渲染时第 `n` 帧必须能独立
- 渲染;
-- 它必须被归一化为精确逐帧计时的视觉素材,或者带类型的动画 Surface;
-- 静态图像即使其文件格式也支持动画,仍然没有时长;
-- 含音频的 MP4 不会因此就让那段音频可听;
-- 带 alpha 的视觉素材必须保留显式的 alpha 合同。
-
-多个 Layer 可以引用同一个 Artifact。图和 Artifact 收集器会对字节做去重;渲染器仍然独立求值每一个被声明的 sample。
-
-## 4. The orthogonal item model
-
-从概念上说,一个独立 Item 拥有:
-
-```ts
-type MediaItemProgram = {
- readonly id: string;
- readonly temporal: TemporalBinding;
- readonly frame: SpatialFrameInput;
- readonly framePresentation: MediaFramePresentation;
- readonly layers: readonly MediaLayerProgram[];
- readonly motion: MediaLifecycleMotion;
- readonly stacking: AbsoluteStackingIntent;
- readonly sourceAudio?: MediaSourceAudioProjection;
-};
-```
-
-各个正交轴是:
-
-| 轴 | 负责 | 不负责 |
-|---|---|---|
-| source | 精确的静态/定时/Surface 素材 | Provider 或素材发现 |
-| temporal | occurrence 展开与 item 窗口 | 素材播放 |
-| frame | Placement Frame | 可见 Paint 或内容宽高比 |
-| layers | 有序的 Paint/媒体 sample | Track 相对的 z 上下文 |
-| content fit | 每个 sample 一个 Content Frame | zoom/pan 或播放 |
-| occupancy | 源裁剪与源时间映射 | 目标窗口 |
-| frame presentation | 裁剪、圆角、描边、阴影、内边距 | 几何构造 |
-| lifecycle motion | 自有像素的进入、持续与退出 | member 之间的 handoff |
-| sampling motion | 单个媒体 Layer 内部的 pan/zoom/旋转 | Placement Frame 的运动 |
-| stacking | 绝对 Present 位置 | Track family 优先级 |
-| source audio | 显式投影某一个被选中 Layer 的音频 | 自动的容器音频 |
-
-这些都是包内的 Program 事实。它们不会变成 Core 或终端 Track 合同上的字段。
-
-## 5. Illustrative author Surface
-
-最终的 Surface 必须优先使用具名 Frame 和 SVS Recipe,而不是大段内联参数堆。一种可能的写法是:
-
-```xml
-<media:Track id="editorial-media" map={timing.map} space={speech.space} canvas={vertical}>
- <media:Item
- id="full-cutaway"
- video={demo.video}
- during={story.selection.demo}
- frame={vertical}
- appearance={studio.media.full-cutaway}
- />
-
- <media:Item
- id="corner-loop"
- surface={reaction.surface}
- at={story.moment.reaction}
- for="2s"
- frame={layout.top-right-sticker}
- appearance={studio.media.corner-loop}
- />
-
- <media:Sequence
- id="steps"
- frame={layout.lower-card}
- appearance={studio.media.evidence-sequence}
- until={story.selection.explanation.end}
- >
- <media:Member id="step-1" image={step1.image} extent={step1.extent} at={story.moment.step1}/>
- <media:Member id="step-2" image={step2.image} extent={step2.extent} at={story.moment.step2}/>
- <media:Member id="step-3" video={step3.video} at={story.moment.step3}/>
- <media:Handoff from="step-1" transition={studio.media.transitions.push-left}/>
- <media:Handoff from="step-2" transition={studio.media.transitions.crossfade}/>
- </media:Sequence>
-
-</media:Track>
-```
-
-这段语法只是示意。规范性的要点是:
-
-- 声明与可复用外观留在使用点之外;
-- 每一个 source、Frame、Selection 和 Moment 都是显式引用;
-- `image`、`video`、`media` 和 `surface` 表示互斥的 source 形态;绝不会去猜一个通用 Blob 到底是静态还是
- 动态素材;
-- `Item` 与 `Sequence` 在作者视角下是明显不同的语义;
-- `during`、`at`、`for`、`until` 这类简写全部下降到同一套 Temporal 代数;
-- 任何 Runtime 都不参与决定某个东西是插叙、卡片还是贴纸。
-
-### SVML structure and SVS parameter bundles
-
-SVML 拥有拓扑:source、Frame、时间引用、Item/Sequence 成员关系、Handoff 边、标签和声音输入。
-SVS 拥有具名的可复用参数包:Content Fit、frame Paint、裁剪、运动参数和转场外观。
-
-SVS Recipe 不能隐藏一条 source 边,不能创建 member,不能选择 Provider,也不能让两个独立的 Item 变成一个 Sequence。把一个转场 Recipe 应用到整个 Sequence,会在解析后的 Program 里把这个选择物化到每一条被声明的 Handoff 上。因此 Recipe 既保持简洁,又不会退化成第二套图语言。
-
-### Small framed inset with a smooth rise
-
-小画框不是另一个组件。它就是一个普通 Item,只不过它显式的 Placement Frame 比 Canvas 小。使用点的结构依然可读:
-
-```svml
-<space:AnchoredFrame id="product-inset" within={safe}
- x="50%" y="78%" width="82%" height="28%" anchor="center"/>
-
-<media:Track id="proof" map={timing.map} space={speech.space} canvas={vertical}>
- <media:Item
- id="product-proof"
- video={product.video}
- during={story.selection.proof}
- frame={product-inset}
- appearance={studio.media.fuzzy-card}
- motion={studio.motion.smooth-rise}
- />
-</media:Track>
-```
-
-被引用的 Recipe 在概念上解析出彼此独立的参数,例如:
-
-```text
-media.fuzzy-card
- content fit contain or cover
- frame fill explicit color/gradient/transparent Paint
- content padding exact pixels
- frame clip rounded rectangle or owned shape
- edge treatment border/shadow, or an explicit rough-edge layer
-
-motion.smooth-rise
- enter operator translate
- from below the Canvas or an exact positive y offset
- to resolved Placement Frame pose
- duration exact frames
- easing ease-out, or a package-compiled custom frame curve
-```
-
-“下方”必须无歧义。`below-canvas` 由 Canvas 和解析后的 Item 尺寸推导出一个画布外的起始位姿;
-`offset-y: 120px` 则从最终 Frame 的下方开始,未必要离开画面。两者都下降为逐帧精确的局部变换,并且都不改变最终的 Spatial Frame。
-
-普通的柔边或发光边缘下降为自有的 border 加 box shadow/drop shadow。真正不规则的毛边、撕纸边或噪点边不是什么魔法 `border-style`:它是一个显式的局部 Layer,比如一张透明边框 Artifact 或一个确定性的
-`CompositableSurface`。由于该 Layer 会影响图拓扑和 Artifact 收集,它的 source 必须在 SVML 里接上(例如写成 `<media:Layer surface={fuzzyFrame.surface}/>`);SVS 可以给它上样式,但不能隐藏这条 source
-边。外层的生命周期包裹器会带着 frame Paint、边缘 Layer 和内容一起运动。
-
-内置的缓动名称可以直接下降到 Visual IR。更复杂的贝塞尔或弹簧 Recipe 仍归 Media 包管,可以被确定性地采样为按帧寻址的关键帧;这不需要 Core、Composition 或 Runtime 学会一种新的运动类型。
-
-## 6. Independent Item timing
-
-Item 使用 [`track-authoring.md`](./track-authoring.md) 中的通用流水线:
-
-1. 定位 Selection、Moment 或 Program 点;
-2. 展开 `one` 或 `each` occurrence;
-3. 为每个 occurrence 投影一个目标窗口;
-4. 与 ProgramSpace 求交并校验;
-5. 在该 item 窗口内求值每一个 Layer 的源裁剪与 occupancy;
-6. 求值生命周期运动与采样运动;
-7. 产出自包含的 Present。
-
-独立的 Item 保持独立。它们的窗口可以重叠,并按各自显式的绝对 stacking key 合成。Program 顺序不会让它们互斥,不会裁掉优先级更低的 item,不会自动缝合空隙,也不会推断出一个 Sequence。
-
-全屏 B-roll 不过是一个 Placement Frame 等于 Canvas、前景 Layer 通常用 `cover` 的 Item。小图、GIF 或视频只是同一个 Item 配上更小的 Frame。“不要挡住人脸”不是 Media 的某个 mode:作者要么消费一个合适的具名 Frame,要么由上游具备主体感知能力的布局组件显式产出一个。
-
-## 7. Ordered local layers
-
-一个 Item 在一个 Placement Frame 内拥有一份有序的局部 Layer 列表:
-
-```ts
-type MediaLayerProgram =
- | MediaFramePaintLayer
- | MediaSampleLayer;
-
-type MediaSampleLayer = {
- readonly id: string;
- readonly source: MediaSourceInput;
- readonly fit: ContentFit;
- readonly trim?: VisualSourceTrim;
- readonly occupancy: VisualOccupancy;
- readonly appearance: MediaSampleAppearance;
- readonly samplingMotion?: MediaSamplingMotion;
-};
-```
-
-[`spatial-layout.md`](./spatial-layout.md) 中列出的全部衬底情形,都只是普通的 Layer 列表:
-
-- 透明:没有衬底 Layer;
-- 纯色或渐变:一个 frame Paint Layer;
-- 自身模糊:同一个 source 进入一个 cover/模糊 Layer 和一个前景 Layer;
-- 替代衬底:由另一个显式 source 提供衬底 Layer;
-- 装饰卡片:Paint、border/阴影/内边距,再加一个或多个媒体 Layer。
-
-每个媒体 Layer 都由自己的固有尺寸和 `ContentFit` 推导出自己的 Content Frame。不存在特殊的前景/背景配对,也不存在唯一的共享内框。
-
-Frame 级裁剪是显式的:不裁、裁到 Placement Frame,或裁到自有形状。圆角、border、阴影、内边距和
-frame Paint 都属于表现层。它们不能改变 Spatial 几何,也不能采样另一个 Track。
-
-## 8. Source-time occupancy
-
-静态素材没有固有播放时长,在其 Layer 处于活动状态的整个区间内持续绘制。定时视频、动图或动画 Surface
-先执行显式 trim,再套用通用的视觉 occupancy 法则:
-
-```text
-once/start play from the source head, then disappear or truncate
-once/end align the source tail to the window end
-hold/start play from the head, then hold the last visual frame
-hold/end hold the first visual frame, then play to the end
-loop/start loop from source-head phase
-loop/end choose phase so the source tail meets the window end
-stretch map the complete effective source interval onto the window
-```
-
-不存在 `auto`、`native`、`finish`、`freeze` 或 `phased` 这类作者枚举值。图像不会被塞一个假的播放模式。源 trim 在 occupancy 之前执行。空间上的 contain/cover 与时间上的 stretch 互不相干。
-
-即使某个 `once` Layer 提前结束,Item 的目标窗口依然有效。其他 Paint 或媒体 Layer 可以继续存在。若作者希望整个 item 保持可见,就选 `hold`、补一个常驻 Layer,或者直接写一个更短的目标窗口;编译器绝不会靠猜哪个 Layer 是主 Layer 来收缩 Item。
-
-## 9. Media motion has four separate channels
-
-旧的 B-roll 行为拆成四条由包自有、彼此显式分离的通道:
-
-```ts
-type MediaLifecycleMotion = {
- readonly enter?: MediaEdgeMotion;
- readonly sustain?: readonly MediaSustainMotion[];
- readonly exit?: MediaEdgeMotion;
-};
-
-type MediaSamplingMotion = {
- readonly keyframes: readonly MediaSamplingKeyframe[];
-};
-```
-
-### Entry motion
-
-作用于一个 Item 或 Group 在其可见包络起始处所拥有的全部像素。合理的带类型 Recipe 包括淡入、方向滑入、缩放、弹出、回弹、局部模糊显现、擦除、翻转和旋转。
-
-### Sustain motion
-
-作用于可见包络期间。漂浮、呼吸、脉冲、摇摆、抖动和漂移都是合理的算子。被声明的有序列表按确定性方式求值;多个变换不会悄悄互相覆盖。
-
-### Exit motion
-
-作用于包络结束处的全部自有像素。它有自己的算子和时长;它不会自动成为 entry 的逆过程。
-
-### Sampling motion
-
-作用于某一个媒体 Layer 已完成 fit 的 Content Frame。归一化关键帧可以驱动缩放、源点/内容点位移和旋转。
-Ken Burns 是一个采样运动 Recipe,既不是 Base 特效,也不是 occupancy 模式。
-
-即使 entry 与 exit 重叠,它们各自被声明的时长也保持不变。包在每一帧同时求值两者,并确定性地合成它们彼此独立的 opacity、transform、filter 和 clip 通道。它绝不会按比例压缩任何一侧,也绝不会把 Item 跑两遍。可见的 Item 窗口仍然是最终的裁剪边界。
-
-变换栈是固定的,因此各通道不会互相覆盖:
-
-```text
-absolute Present stacking
-└─ fixed Placement Frame
- └─ group/item entry + exit wrapper
- └─ ordered sustain wrappers
- └─ member handoff wrapper (Sequence only)
- └─ frame Paint and clip
- └─ base fitted Content Frame
- └─ layer sampling motion
- └─ media pixels
-```
-
-下降器可以在代数上合并矩阵,也可以产出嵌套的 Visual IR 元素,但顺序和像素必须完全一致。这些是 Media
-包的合同,不是某个通用的公开 `effects` 参数袋。
-
-## 10. Sequence: explicit member replacement
-
-一个 Sequence 拥有两个或更多嵌套的 Member,它们共享一个 Placement Frame 和一个外层生命周期。它必须被显式创建;相邻的 Item、重叠的窗口或某个转场 Recipe 都不会隐含成员关系。
-
-每个 Member 拥有:
-
-- 稳定的作者身份;
-- 显式的媒体 Layer 以及逐 Layer 的 fit/occupancy/采样运动;
-- 恰好一个由 Moment 或某个 Selection 边界推导出的激活点;
-- 没有独立的外层 entry 或 exit 运动。
-
-Sequence 有一个显式的终止点。对于激活帧 `p1..pN` 和终止点 `T`:
-
-```text
-p1 < p2 < ... < pN < T
-
-logical member 1 phase [p1, p2)
-logical member 2 phase [p2, p3)
-...
-logical member N phase [pN, T)
-```
-
-源码中的 occurrence 顺序具有权威性。编译器绝不按物理时间排序。缺失、同帧、逆序或超出 Program 范围的激活点,会让整个 Sequence 原子性失败。
-
-每一对相邻成员都拥有一条显式的 Handoff。不存在只活在编辑器或 Runtime 里的转场默认值;“应用到全部”会把同一个被选中的转场物化到每一条边上。
-
-### Pair transition
-
-官方的初始集合是:
-
-- cut;
-- crossfade;
-- push;
-- wipe;
-- cover;
-- page turn。
-
-一个转场拥有算子身份、精确时长、带类型的参数,以及位于 `[0,1]` 的边界位置比例 `r`。给定逻辑边界 `p`
-和时长 `d`:
-
-```text
-handoff.start = p - r × d
-handoff.end = p + (1-r) × d
-```
-
-帧量化保持转场总长度精确不变。相邻的 handoff 窗口不得重叠;非法时长直接失败,而不是造出一个隐式的三源转场。
-
-在 handoff 期间,退出与进入的 Member surface 都由该 Sequence 拥有。它们的媒体 sample 在各自被展开的视觉跨度上连续播放。转场既不改变 ProgramSpace,也不改变上游的 Selection/Moment 点。
-
-Sequence 的 handoff 只能在其共享的 Placement Frame 内部工作。这让 cut、crossfade、push、wipe、cover 和
-page turn 都成为诚实的局部关系。针对 `composite_below` 的外层 handoff 被刻意废弃:它会读取并改动无关的下层 Track。
-
-Sequence 自身可以使用普通的 entry/exit 运动。在下层 Track 之上做全屏淡出是合法的,因为改变的只是
-Sequence 的不透明度;而把下层 Track 推走或翻页则不合法。
-
-## 11. Audio is explicit and projected separately
-
-视觉素材默认静音,除非作者显式地从恰好一个具名媒体 Layer 中选取源音频。这是 item 级别的,因此前景与它自身的模糊副本不会意外地把同一个源混两遍。
-
-```ts
-type MediaSourceAudioProjection = {
- readonly fromLayer: string;
- readonly gain: number;
-};
-```
-
-音频投影复用被选中 Layer 的精确 trim、播放映射和已解析的调度。Sequence 的每条边各自拥有自己的音频关系:
-
-- `cut` 在逻辑激活点切换可听性,即使视觉 surface 存在重叠;
-- `crossfade` 使用显式声明的音频 handoff 窗口;
-- 视觉转场绝不会悄悄替你选一个音频转场。
-
-entry/exit 或 handoff 的音效是显式的音频 Artifact 输入,带显式增益。它们在解析出的精确 entry、exit 或边界点上下降。它们不是包级全局文件名、不是可变的音效库默认值,也不是隐藏的通道音效。
-
-编译器不会返回一个同时包含两个终端 Track 的原子 `MediaProduct`。它解析出一个包自有的 Media Program,并提供彼此独立的确定性投影 Operation:
-
-```text
-Resolved Media Program ─┬─> project visual ─> VisualTrack
- └─> project audio ─> AudioTrack
-```
-
-因此,任一终端输出都可以被 Run Graph 独立地定向、替换或满足。当两者同时被需求时,它们仍通过一条显式的图边共享同一份已解析调度,而不是复制两份时序元数据。
-
-一个纯视觉 Item 不会仅因为它的源容器里带音频流,就去要求音频归一化或渲染。
-
-## 12. Stacking and flat Tracks
-
-Media Track 自身没有 z-index。每一个解析出的顶层 Item 或 Sequence surface,以及任何可被单独穿插的
-member/Layer,都会产出绝对的 Present stacking key。
-
-frame Paint 与紧耦合的媒体 sample 可以留在同一棵 Present 局部元素树里。当对等 Track 必须能出现在某块板、卡片、图标或 Layer 之间时,包会产出带独立绝对 key 的独立 Present。Track 归属关系绝不创建 stacking
-上下文。
-
-Sequence handoff 内部的临时顺序由包自有,且只在该关系的那些帧内存在。它不会预留一条全局 z 带。
-
-## 13. Speech Spine's restricted Media projection
-
-Speech Spine 不应再维护第二套临时的媒体渲染器。它的视觉投影使用同一套 Media 下降实现,只是生成的
-Program 被刻意限制:
-
-```text
-one normalized speech-bearing Take per Segment; its visual stream is optional
-exact already-established Segment frame span
-explicit authored Spatial Frame
-one foreground layer
-explicit fit and absolute stacking order chosen by the author
-muted visual
-cut between contiguous takes
-no entry, sustain or exit motion
-no Sequence state
-no source-audio projection
-```
-
-Speech 的作者 Surface 接受原始 `video=`,并在装配前展开同一套通用的检查与归一化图,按 Spine 显式的帧率选取主运动视频加默认音频。原始 `audio=` 展开为音频权威的归一化图,不贡献任何视觉 Present;它绝不凭空造出黑场素材。已准备好的 `media=` 仍是直接的精确输入。视觉 Take 继承 Spine 显式的 `visual-frame`、
-`visual-appearance` 和 `visual-z`,并可在这几个受限的轴上做逐 Take 覆盖。装配之后,Speech 音频仍走独立的规范路径 `SpeechAudioBasis -> AudioTrack` 投影;受限的视觉下降器本身保持静音,不会去重新发现容器音频。
-
-这可以是对聚焦的 Media 下降辅助函数的代码依赖。当没有任何外部组件消费那个中间值时,它并不需要暴露一个公开的 `MediaProgram` 图 Type。若日后真的出现图消费者,包可以暴露自己的带版本已解析 Type;Core 仍然不会注册 Media 语义。
-
-作者导入的是 Speech Spine,其 manifest 声明了自身的包依赖。作者不需要仅仅因为实现复用了 Media 下降,就再导入第二个解析器。
-
-普通 Media Item 可以直接消费 `during={story.segment.answer}`。这不是隐式 Selection:Temporal 解析的是
-Segment 早已存在的结构性起止锚点。当作者的真实意图是 Selection 和 Moment 时,同一个 Item 仍然可以消费它们。
-
-## 14. Legacy audit: retained and retired
-
-保留:
-
-- 通过统一的 Item 模型表达全画幅插叙、分屏和角落叠加;
-- 静态图像、类 GIF 动画和视频;
-- 彼此独立且可重叠的媒体呈现;
-- 显式的 contain/cover/焦点采样,以及前景/自身模糊衬底;
-- 源 trim 与 once/hold/loop/stretch occupancy;
-- entry、exit、持续的外层运动和内部采样运动;
-- 显式的 Sequence 成员关系与共享 surface 的成对转场;
-- 提示位置比例,以及跨转场窗口的连续视觉采样;
-- 显式的源音频与转场/边缘音效;
-- 稳定的作者身份与确定性的声明顺序。
-
-废弃:
-
-- 独立的 B-roll 终端 family;
-- 会改变合法端口或设置的 `mode`;
-- 用 URL 扩展名猜测 image/video/GIF 行为;
-- Track 级 z-index 与隐式 item 优先级;
-- `auto` 播放、`finish`、`freeze`、`phased` 以及渲染器兜底修复;
-- 自动排序或 `linkNext` 推断;
-- 用同一组 `focalX/focalY` 同时表示源点和目标点;
-- 用固定的前景/背景字段取代有序的局部 Layer;
-- 用一个 CSS transform 槽位让各运动通道互相覆盖;
-- 对重叠的 entry/exit 时长按比例压缩;
-- 通道级全局音效与自动选中的音频;
-- 外层 `composite_below` 转场以及一切 Base FX 依赖;
-- 把 VisualTrack 与 AudioTrack 的满足耦合在一起的打包式 B-roll Product。
-
-旧的 Deck 行为在 [`deck-track.md`](./deck-track.md) 中独立迁移;把它从 Media Track 移除并不等于删掉这项能力。
-
-## 15. Package and implementation shape
-
-预期的包关系是:
-
-```text
-@narratage/media-track
-├─ author Surfaces: Track, Item, Sequence, Member, Handoff
-├─ package-owned Program validators
-├─ pure Item window/layer/motion lowering
-├─ pure Sequence schedule + pair transition lowering
-├─ VisualTrack projection
-└─ AudioTrack projection
-
-depends on
-├─ @narratage/media
-├─ @narratage/temporal
-├─ @narratage/spatial
-├─ @narratage/svs
-└─ @narratage/composition
-```
-
-包内部可以使用聚焦的文件划分,或未来仅供实现使用的库。但它不得造出一个被 Text、Caption、Ranking 和
-Media 共用的大一统创作库,也不得在新增一个媒体 Recipe、运动算子或组件包时要求 Core 发版。
-
-原先的 `@narratage/broll` 包在迁移期间用作回归见证,现已废弃。由于项目处于预发布阶段,不保留任何兼容包。“B-roll” 仍是描述某个 Item 或 Sequence 的有用剪辑术语,而不是终端 Type 或作者包。源素材与生成文件保持原样,不受影响。
-
-## 16. Implementation order
-
-只有在共享的 Temporal 与 Spatial 切片就位之后才实现:
-
-1. **已实现:** 引入 `@narratage/media-track`,先支持一个静态图像 Item,其 BlobArtifact、
- IntrinsicExtent、SpatialFrame、ContentFit 和 ProgramSpace 是彼此独立的语义输入;
-2. **已实现:** 加入定时素材的源 trim 以及全部 occupancy/对齐策略;
-3. **已实现:** 加入有序的 Paint/媒体 Layer、裁剪和双 Frame 的 fit 模型;
-4. **已实现:** 加入生命周期运动与采样运动,并固定变换栈;
-5. **已实现:** 加入显式的 Sequence 调度与成对转场;
-6. **已实现:** 加入独立的源音频与音效投影;
-7. **已实现:** 把 Speech Spine 的视觉投影接到受限下降器上;
-8. **已实现:** 迁移仓库中已签入的 B-roll 示例并废弃旧包;
-9. **已实现:** 用结构性证据、真实媒体证据和分区浏览器证据闭合包内验收矩阵。
- [`track-expressiveness.md`](./track-expressiveness.md) 中独立的共享 Track/Visual IR 关卡同样通过。
-10. **已实现:** 让 `video=` 按已连接的 ProgramSpace 帧率下降到共享的 Media Pipeline,同时保留显式的
- `media=` 作为“已准备素材”的逃生口。
-
-其中没有任何一步需要改动 Core、Runtime、队列或 Provider。Surface 物化可以使用某个已显式注册的 Provider
-能力,但普通的 Item/Sequence 下降是确定性的本地编译。
-
-## 17. Acceptance matrix
-
-### Material
-
-- 不透明与带 alpha 的静态图像;
-- 竖版、横版和正方形图像;
-- 生成的/原始视频,在有和没有显式选中音频两种情况下自动归一化;
-- 已显式准备好的归一化视频直接接入,不再做第二次归一化;
-- 动图 GIF/WebP 归一化为精确逐帧计时;
-- 静态与动画的 Compositable Surface;
-- 重复使用同一个 Artifact,不产生重复存储或重复源音频。
-
-### Item
-
-- `one` 下的 Program/绝对窗口,以及 `one` 和 `each` 下的 Selection/Moment 窗口;
-- 位于互不相关的绝对 stacking key 上、彼此独立且重叠的 Item;
-- 全部源 trim 情形,以及更短/相等/更长的 occupancy 情形;
-- 全 Canvas、分屏、下方卡片、角落和画布外 Frame;
-- 透明、纯色、渐变、自身模糊和替代 source 的衬底 Layer;
-- 每一种 Content Fit,以及内容焦点与 frame 焦点不一致的情形;
-- entry、sustain、exit 与采样运动,分别使用以及组合使用;
-- 对重叠的 entry/exit 时长做确定性合成,不做隐藏的重新计时。
-
-### Sequence
-
-- 图-图、视频-视频、图-视频以及 alpha Surface 之间的 handoff;
-- 严格的声明激活顺序与稳定的 member 身份;
-- cut、crossfade、push、wipe、cover 和 page turn;
-- start/center/end 三种边界位置比例;
-- 跨 handoff 重叠区间的连续源采样;
-- 转场时长小于、等于和大于相邻逻辑相位间距的情形,前提是显式 Sequence 包络与“无三源”法则仍然成立;
- 非法的包络/重叠必须失败;
-- 缺失、重复、逆序和同帧激活的失败情形;
-- 不存在隐式的三成员重叠,也不存在下层合成输入;
-- 组的 entry/exit 只作用于自有像素。
-
-### Audio and execution
-
-- 纯视觉 source 不产生任何音频工作需求;
-- 恰好一个被显式选中的 source Layer 贡献音频;
-- 硬切与交叉淡化两种 Sequence 音频 handoff;
-- 同一边界上同时发生的 exit/entry 音效混合;
-- 独立的 BGM 保持不变;
-- 独立地定向/替换视觉与音频投影;
-- 本地与远程执行下产出完全一致的 Visual Track/Audio Track 计划;
-- 任意帧分区渲染出字节级一致的确定性帧。
-
-### Architecture
-
-- 安装另一个 Media 组件不会改动 Core、Film、Composition 或 HyperFrames 的任何 family 注册表;
-- 所有 source、时间、空间、声音和标签依赖都是图边;
-- 任何 Track 都不包含 provider、路径、Narrative 摘要或 source family 元数据;
-- 没有任何 operation 采样已累积的下层合成结果;
-- Speech Spine 的受限投影与普通 Item 遵循同一套下降法则。
-
-### Executable evidence
-
-上面的矩阵由可执行证据闭合,而不是靠包状态的文字描述:
-
-- `packages/media-track/test/media-track.test.ts` 覆盖每一种 Item 绑定、严格的 occurrence 展开、全部
- 已记录的 Frame 与 Content Fit、有序的 Paint/sample/衬底 Layer、trim 与 occupancy、生命周期/sustain/采样
- 运动、每一个 Handoff 算子与比例、精确的源音频/音效投影、非法 Sequence 的拒绝,以及显式的作者图边;
-- `packages/provider-media-local/test/provider.test.ts` 对带音频和不带音频的运动视频、附带图片、源 A/V
- 偏移、动图 GIF、动图 WebP、精确音频渲染和最终混流跑真实的 FFmpeg 检查与归一化;
-- `packages/media-execution/test/webp.test.ts` 单独证明本地与 Lambda 执行共用的动图 WebP 直通 alpha 混合与
- dispose 行为;
-- `packages/provider-media-aws-lambda/test/provider.test.ts` 证明远程执行收到的是完全相同的
- `AudioProgramPlan`;本地与 Lambda 媒体环境都执行 `@narratage/media-execution`,而不是各自拥有另一套
- 媒体语义;
-- `packages/speech-spine/test/surface.test.ts` 和 `packages/speech-basis/test/product.test.ts` 证明受限的
- Speech 投影、单一共享的生成 Product,以及视觉/音频 Candidate 的独立满足;
-- `packages/hyperframes/test/browser-visual.test.ts` 通过真实 Chromium 渲染不透明图像、直通 alpha Surface、
- 双 Frame fit、局部运动和一次 Sequence handoff,然后跨不同 worker 分区比对每一帧解码后的 RGBA;
-- 包 manifest 中不含任何 Media family 注册表,公开的值类型中也不含隐藏血缘或 Provider 元数据。
-
-因此,本地包作为构建于冻结的仓库内部 `svml.visual-track@1` / `svml.visual-ir@1` 窄腰之上的预发布 Media
-实现已经完整。这并不意味着 Media 作者 Surface 已被发布或被单独冻结。
diff --git a/docs/zh/guide/spec/ranking-track.md b/docs/zh/guide/spec/ranking-track.md
deleted file mode 100644
index 57bd697b..00000000
--- a/docs/zh/guide/spec/ranking-track.md
+++ /dev/null
@@ -1,257 +0,0 @@
----
-title: SVML Ranking Track 迁移
-description: 官方 Ranking 迁移的已实现可执行权威,包含全部四个作者组件。
----
-
-# SVML Ranking Track 迁移
-
-状态:官方 Ranking 迁移的已实现可执行权威。四个作者组件、它们共享的触发式调度、可选的同级 AudioTrack
-lowering、作者 Surface 以及真实浏览器证据都在 `@narratage/ranking` 中执行。它不是对历史 node 格式的兼容承诺,也不是冻结的公开 ABI。
-
-## 1. Ranking 的含义
-
-Ranking 是一个视频领域的作者组件,它把一个有序集合呈现为渐进的视觉状态。它不是 Core 概念、不是通用的
-Track 模式、不是 Provider,也不是渲染器特性。
-
-旧系统把四种本质不同的行为塞在同一个 `ranking_track` 节点和一个 `rankingType` 开关背后。新系统使用一个物理包 `@narratage/ranking@1`,但暴露四个互相独立的作者组件:
-
-- `ranking:TierBoard` —— 把已揭示的条目放入具名的 tier 行;
-- `ranking:Column` —— 在舞台上展示一个当前条目,然后让它落位到带编号的列中;
-- `ranking:TopThree` —— 渐进揭示最多三个条目,并强调当前条目;
-- `ranking:TypewriterList` —— 在常驻的纸张 / 列表表面上逐字打出有序文本行。
-
-一个包是共享校验、调度构造、布局工具和绘制原语的正确单位。四个组件是作者语义的正确单位。不存在公开的
-`rankingType`、模式选择输入、动态端口注册表或 Runtime 渲染器注册表。
-
-第三方可以发布另一个 Ranking 包,无需在 Core 注册类型,也无需修改本包。安装并 import 该包后,它会通过普通的模块机制贡献自己的 Surface、Type、validator 和 fragment 编译器。
-
-## 2. 来自旧实现的证据
-
-被审阅的旧项目清单中包含 21 个已迁移的 Ranking 节点:
-
-| 历史变体 | 审阅到的用例数 | 观察到的条目数 |
-|---|---:|---|
-| Tier | 10 | 5–6 |
-| Typewriter | 5 | 5 |
-| Column | 3 | 4–5 |
-| Top Three | 3 | 3 |
-
-这份清单证明了四种语义都真实存在,但并不证明它们的旧表示应当延续。旧实现存在如下结构性问题:
-
-1. 同一个 UI / node 身份通过切换模式并断开其动态端口来改变语义;
-2. 同一对条目 `start/end` 在不同渲染器下分别意味着可见性、动画阶段、后继交接,或者什么也不意味着;
-3. Top Three 忽略条目结束时间,而 Typewriter 只用它们的并集来判断纸张是否已挂载;
-4. 渲染器检视所有兄弟片段、对其排序或计数,并在通用 Track Program 已经丢失这些语义之后又把组状态制造
- 出来;
-5. 一个全局 z-index 使看板与各个图标无法与同级 Track 交错;
-6. 裸 URL、环境字体名、Remotion 组件 id 以及三份互为镜像的注册表把环境 / 渲染器细节泄漏进组件;
-7. 过短的窗口可能悄悄压掉运动音效,而不是把视觉与声音阶段一起适配到同一份有限的共享调度中。
-
-迁移保留的是有用的视觉能力,而不是这些偶然形成的语义。
-
-## 3. 面向作者的形态
-
-可执行的作者 Surface 是:
-
-```svml
-<import as="ranking" from="@narratage/ranking@1"/>
-<import as="fonts" from="@narratage/fonts-open@1"/>
-<import as="studio" source="./studio.svs"/>
-
-<fonts:Stack id="ranking-fonts" family="inter" weight="700" style="normal"/>
-
-<ranking:ColumnStyle id="column-style"
- recipe={studio.ranking.column}
- font={ranking-fonts}/>
-
-<ranking:Column id="tools"
- space={speech.space}
- map={timing.map}
- during={story.selection.ranking-board}
- triggers={story.moment.next-rank}
- terminal={story.moment.ranking-complete}
- style={column-style}>
- <ranking:ColumnItem label="Fourth" icon={fourth.image}/>
- <ranking:ColumnItem label="Third" icon={third.image}/>
- <ranking:ColumnItem label="Second" icon={second.image}/>
- <ranking:ColumnItem label="First" icon={first.image}/>
-</ranking:Column>
-```
-
-`during` 提供显式的外层窗口。`triggers` 提供一个有序的 Moment occurrence 集合。`terminal` 是最后一个活动阶段的显式终点;它可以早于外层终点,从而让完全落位后的结果继续可见。组件要求 trigger occurrence 的数量等于子条目的数量,并按作者书写顺序一一对应。
-
-外层 Selection 与 terminal Moment 都使用严格的 `one` 语义。trigger Moment 使用严格的 `each` 语义。外层 /
-terminal 出现零次或多次、trigger 出现零次,以及任何 trigger 与条目的基数不匹配,都是编译错误;组件绝不会自己挑一个方便的 occurrence。
-
-这种按位置配对由这四个组件声明,不是通用的 Core 行为。另一种第三方 Surface 可以给每个条目单独的 Moment
-边,并 lower 到同一套包内事实。
-
-Style 声明与使用保持分离。SVS 提供具名的通用 Recipe 数据;每个变体专属的 `*Style` 组件拥有各自允许的键,并把 Recipe 加上显式的字体边编译成一个带类型、由包拥有的 Style。诸如 tier 行、标签、图标和入场行为这类语义化的条目值不是绘制属性,不属于 SVS。
-
-## 4. 组能看到兄弟,渲染出的 Track 不能
-
-渐进式 ranking 不能通过把每个子项当作孤立的 Track item 来编译。行位置、后继交接、活动条目状态和已落位的前缀都依赖这个有序的组。但这不意味着条目在渲染时互相检视。
-
-作者组件首先把完整的子项列表和 trigger 集合编译成一份包内私有的调度:
-
-```ts
-type RankingSchedule = {
- readonly outer: FrameSpan;
- readonly terminalFrame: number;
- readonly entries: readonly {
- readonly itemId: string;
- readonly triggerOccurrenceId: string;
- readonly triggerFrame: number;
- readonly stage: FrameSpan;
- readonly cumulative: FrameSpan;
- }[];
-};
-```
-
-这是确定性 Operation 之间一个显式的 Graph 值,不是夹带在图像、Track 或媒体 Artifact 里的元数据。它由
-Ranking 包拥有并版本化,不会加入 Core、Composition 或共享的 Track 窄腰。
-
-```text
-Moment set + SemanticMap + ProgramSpace + outer/terminal projection + ordered item ids
- │
- ▼
- RankingSchedule
- │ │
- Style + content ───┘ └── SFX clips + gains
- │ │
- ▼ ▼
- VisualTrack AudioTrack
-```
-
-调度 Operation 保留源顺序和稳定的 occurrence 身份,绝不按物理时间排序。帧量化之后它要求:
-
-```text
-outer.start <= p1 < p2 < ... < pN < terminal <= outer.end
-```
-
-任何非法或帧相等的序列都会整体失败。视觉与可选音频的 lowerer 消费同一份调度,因此运动与声音不会漂移,同时它们仍然是 Run Graph 中可以各自独立被 demand 的分支。
-
-## 5. 共同的阶段模型
-
-对外层窗口 `[B, O)`、trigger `p1 ... pN` 和 terminal `T`,条目 `i` 得到:
-
-```text
-active stage [p_i, p_(i+1)) or [p_N, T)
-cumulative life [p_i, O)
-settled life [stage.end, O)
-```
-
-`[B, p1)` 可以显示作者显式声明的空看板,但没有隐含的活动条目。`[T, O)` 是可选的完全落位后缀,没有活动条目。这些事实是由包拥有的兄弟排序,建立在 [`track-authoring.md`](./track-authoring.md) 的共享时间投影代数之上;它们不是一个新的 `Stage` 合同。
-
-### TierBoard
-
-- 看板占据 `[B, O)`;
-- 每个条目从它的 trigger 起累积可见直到 `O`;
-- `entry="direct"` 在 trigger 之后直接动画进入它最终的 tier 单元格;
-- `entry="stage"` 在共享的 stage 点出现,在其活动阶段内停留于此,并在该阶段结束前移动到最终单元格,之后
- 保持落位状态;
-- 条目拥有必填的图像和 tier 行 id;行内的源顺序决定它的单元格;
-- 不可能完成的出现 / 移动时长是编译错误,绝不会被悄悄缩短或省略。
-
-### Column
-
-- 带编号的看板占据 `[B, O)`;
-- 每个活动阶段内恰好有一个条目占据大舞台;
-- 该阶段结束时,这个条目成为已落位列的一部分,直到 `O`;
-- 标签必填,图像可选,名次即作者书写的条目顺序;
-- 在一份合法的调度之后,当前条目重叠在结构上不可能发生。
-
-### TopThree
-
-- 最多接受三个条目;
-- 每个条目从它的 trigger 起累积揭示直到 `O`;
-- 只有对应当前活动阶段的条目会得到活动光环 / 强调运动;
-- `[T, O)` 展示所有已揭示条目且没有活动强调;
-- 标签必填,图像可选,槽位即作者书写的条目顺序。
-
-### TypewriterList
-
-- 纸张占据 `[B, O)`,空白行可以在其 trigger 之前就在结构上预留;
-- 第 `i` 行在 `p_i` 开始打字,并保持可见直到 `O`;
-- 打字时长由 Unicode 字素簇数量和 Style 的速度确定性地推导;
-- 打字以及可选的获胜标记必须放得进该条目的活动阶段,否则编译失败;
-- 标题、可选的显式强调区间和各行字符串都是作者内容,不是 LLM 字段。
-
-Typewriter 之所以留在 Ranking 包里,是因为尽管它大量使用文本排版,它的作者语义仍然是一个有序的渐进
-ranking。它可以与未来的 Text 实现共享底层的精确字体 / 排版工具,但不得依赖通用 Text Track Program,也不得冒充成它。
-
-## 6. 空间布局与 stacking
-
-每个组件拥有一个外层放置框和它自己的内部布局算法。Tier 单元格、Column 行或 Top Three 槽位不是通用的逐条目空间定位器。允许在这些子项上传入任意空间输入会摧毁组件的语义。
-
-放置与 stacking 保持独立:
-
-- 外层框决定看板、舞台和条目的坐标系;
-- Style 解析出看板的绝对 stack 顺序,以及舞台 / 条目的默认值;
-- 每个子项都可以显式覆盖自己的绝对条目 stack 顺序;
-- 在需要独立交错时,看板、舞台装饰和每个条目会 lower 成各自独立的 `VisualPresent` 值;
-- 产出的那一个 `VisualTrack` 不是 stacking context,因此同级 Track 的 Present 可以出现在看板与任意图标
- 之间。
-
-任何包都不应仅仅为了获得多个 z 位置而发出多个 Track。稳定的 `tieBreak` 值来自作者声明的组件 / 条目身份,绝不来自求值顺序。
-
-所有归一化几何都相对于显式的外层框。各变体必需的尺寸参数包括:
-
-- Tier:行定义 / 颜色、看板与标签宽度、stage 点、单元格间距、图标 fit / 圆角以及出现 / 移动时长;
-- Column:看板框、行高 / 行距、stage 点与尺寸、单元格 / 图标几何、名次配色以及出现 / 移动时长;
-- Top Three:中心 / 宽度、标签行、图标尺寸、光环宽度 / 配色以及揭示 / 强调运动;
-- Typewriter:纸张主题 / 绘制、内边距、行距、标题 / 条目排印、旋转、打字速度以及获胜标记的绘制 / 运动。
-
-绘制、几何和运动参数由拥有它们的 `*Style` 组件校验。未知的 Recipe 属性一律失败。不存在任意的 CSS 逃生出口。
-
-## 7. 媒体、字体与声音
-
-图像和声音通过显式的 `BlobArtifact` 边进入。字体通过显式的、内容寻址的字体栈边进入。Ranking 绝不存储
-URL、不要求 Runtime 挑选字体,也不指名某个渲染器组件。
-
-可选音效是普通的作者输入:
-
-- 出现音在条目的 trigger 处开始;
-- 移动音来自视觉 lowerer 所用的同一段已调度的移动阶段;
-- 增益与素材选择位于 Style / 显式声音输入中;
-- 配置了声音就产出同级的 `AudioTrack`;没有声音就不产生被 demand 的音频分支;
-- 声音播放绝不进入 Hyperframes 或某个视觉 Track。
-
-Ranking 自身是确定性的,没有 Need、Provider、队列、凭据或环境适配器。如果输入图像是生成出来的,那次上游生成仍然是另一个 Graph 组件。
-
-## 8. Run Graph 行为
-
-作者图导出普通的逻辑 `visual` 结果,以及在作者声明时导出 `audio` 结果。Target、Candidate 和替换受与其他所有组件相同的 Run Graph 法则约束。Ranking 不新增任何缓存、pin 或预览原语。
-
-替换视觉结果不会改动音频结果。通过普通的显式输出映射,也可以用一个多输出 Candidate 同时替换两个输出。一旦每个被 demand 的原始输出都由选定的 Candidate 满足,demand 分析就会剪掉原来的 Ranking 分支及其时间 /
-媒体祖先。不需要任何特殊的 Ranking 执行规则。
-
-## 9. 迁移顺序与验收
-
-按以下顺序实现:
-
-1. **已实现:** 共享的时间投影与触发式调度法则;
-2. **已实现:** `@narratage/ranking@1`,包含四个相互独立的 Program / Style 和一个私有调度工具;
-3. **已实现:** TierBoard 与 Column,两者共同证明累积语义、独占舞台语义以及独立 stacking;
-4. **已实现:** TopThree;
-5. **已实现:** 使用精确字体 / 排版原语的 TypewriterList;
-6. **已实现:** 从同一份调度产出可选的 AudioTrack lowering;
-7. **已实现:** 真实浏览器的渐进状态见证,以及每个组件一份作者 Surface 图检查。
-
-验收要求:
-
-- trigger 与条目基数、稳定身份以及所有边界失败都有测试覆盖;
-- `T < O` 在画面上可见地证明存在最终的落位后缀;
-- 短阶段在不改变阶段边界的前提下确定性地容纳其局部动画与声音阶段;极短的 Typewriter 阶段可以在同一帧
- 揭示多个字素簇;
-- 看板与逐条目 Present 能在互不相关的绝对 z 位置上与同级 Track 交错;
-- 视觉与声音的事件帧完全一致;
-- 只有声明图像可选的变体才允许缺失可选图像;
-- 字体与媒体都是内容寻址的依赖;
-- 每个组件都能在不改动 Core、Runtime、Composition、Hyperframes 或另一个 Ranking 组件的情况下修改自己的
- Program;
-- 安装第五个组件包不需要改动任何中心家族注册表。
-
-本次迁移有意不保留历史 node JSON、模式切换、旧的动态端口、渲染器 id 或 Remotion 组件。历史项目是证据和视觉参考,不是一门尚未发布的语言的兼容义务。
-
-包的测试矩阵覆盖上述每一条验收项。可选启用的 Chromium 见证会用单 worker 与分区 worker 分别求值全部四个组件,证明可见的 trigger 推进和最终的落位后缀,并且只对跨进程边缘抗锯齿的微观噪声设定容差。逻辑帧状态、几何、stacking、事件帧和作者声明的图边都保持精确。
diff --git a/docs/zh/guide/spec/screen-overlay.md b/docs/zh/guide/spec/screen-overlay.md
deleted file mode 100644
index c7b47547..00000000
--- a/docs/zh/guide/spec/screen-overlay.md
+++ /dev/null
@@ -1,207 +0,0 @@
----
-title: SVML Screen Overlay 作者面
-description: 可选官方 Screen Overlay 包的已实现预发布合同及其自描述 Surface。
----
-
-# SVML Screen Overlay 作者面
-
-状态:可选官方 Screen Overlay 包的已实现预发布合同。它的自描述 Surface、十一个带类型的组件以及顺序 /
-并行浏览器证据均可执行;它还不是冻结的公开 ABI。
-
-## 目的
-
-`@narratage/screen-overlay` 在作者声明的时间窗口内贡献自包含的、通常是全画布的像素。一次 Flash、暗角、扫描线薄纱或确定性漏光,都只是普通的 `VisualTrack`,与 Text、Caption、Media 或 Ranking 完全一样。
-
-它不是调整图层、不是有特权的后合成钩子、不是转场引擎,也不是读取下层 Track 像素的手段。
-
-```text
-Moment / Selection / Program ─> shared temporal projection ─┐
-CanvasSpace ────────────────────────────────────────────────┼─> Overlay Program
-effect parameters / explicit owned assets ─────────────────┘ │
- ▼
- Visual IR elements or owned Surface
- │
- ▼
- VisualTrack
-```
-
-Core、Runtime 和 Composition 都不认识 Screen Overlay。Composition 只校验并排列它那些自包含的 Present。
-
-## 1. 自包含像素法则
-
-每个 overlay 条目都必须仅凭以下内容就能渲染:
-
-- 它自己作者声明的参数;
-- 显式连接的 CanvasSpace 与 ProgramSpace;
-- 该条目拥有的零个或多个显式资产;
-- 它确定性的局部帧号。
-
-它不得采样:
-
-- 位于其下方的累积 Composition;
-- 兄弟 Track;
-- 隐藏的 Base Track;
-- 上一个或下一个场景;
-- 浏览器 backdrop 或环境纹理。
-
-普通的 source-over alpha 合成是允许的:每个带 alpha 的 Track 都是这样抵达画布的。凡是结果依赖未知下层像素的混合模式或滤镜,都不是自包含的。
-
-这条法则意味着旧的、无源的 `backdrop-filter` 版高斯模糊、Color Adjust 和 Zoom Blur 都不是 Screen Overlay。若要诚实地保留这些效果,媒体效果组件必须通过图边接收要变换的精确媒体或可合成 Surface,然后输出它自己变换后的素材或 Track。如果将来某个组件显式地把一组选定的 Track 光栅化成一个 Surface,消费那个 Surface 同样是显式的;Screen Overlay 永远不会获得一个隐秘的“当前帧”输入。
-
-## 2. 作者包,不是算子注册表
-
-官方包可以在同一个命名空间下暴露若干可读的作者组件:
-
-```xml
-<screen:Track id="screen-paint" canvas={vertical} space={speech.space}>
- <screen:Flash at={story.moment.hit} map={timing.map} for="240ms" z="90"
- color="#ffffff" intensity="0.9" attack="2" hold="3" decay="5"/>
- <screen:Vignette during="program" z="20"
- center-x="0.5" center-y="0.5" radius-x="0.82" radius-y="0.68"
- softness="0.3" color="#000000" opacity="0.28"/>
- <screen:Grain during="program" z="70"
- amount="0.12" size="3" chroma="monochrome" motion-rate="0.5" seed="23"/>
-</screen:Track>
-```
-
-这只是示意性的 Surface 语法。编译出的 Program 包含显式的组件身份、参数和时间绑定;它不是一个由 Core 拥有的、字符串化的 `effect="..."` 开关。
-
-对这个小规模的官方集合来说,一个包很方便,但它不是中心注册表。第三方可以发布另一个自描述的包,把新组件
-lower 到同一个 Visual Track 窄腰。只要现有的 Visual IR 或一个带类型的 Surface 能表达出来,新增一个 overlay
-就不得要求改动 Core、Film、Composition 或 HyperFrames。
-
-外观默认值和具名组合属于 `.svs` recipe。包拥有参数校验和确定性 lowering。Runtime Profile 与 Provider 不选择视觉效果。
-
-## 3. Program 的各个轴
-
-每个 overlay 条目都有四个相互独立的作者关注点:
-
-```ts
-type ScreenOverlayItemSpec = {
- readonly id: string;
- readonly content: ScreenOverlayComponent;
- readonly projection: TemporalWindowProjection;
- readonly expansion: OccurrenceExpansion;
- readonly stackingOrder: number;
-};
-```
-
-- `content` 标识实际的视觉组件及其带类型的参数;
-- 组件自有的包络与局部运动参数留在那个带类型的组件内部,而不是塞进一个通用的呈现口袋;
-- `projection` 与 `expansion` 使用共享的 Selection / Moment / Program 代数;
-- `stackingOrder` 结合稳定的作者身份解析为一个普通的绝对 Present 键。
-
-画布几何在这里不是作者选择的 Spatial Frame。Screen Overlay 有意拥有整个连接进来的 CanvasSpace。局部画面的光效、图像、卡片或模糊属于普通的 Media、Text 或组件 Track,使用
-[`spatial-layout.md`](./spatial-layout.md)。
-
-条目默认相互独立。重叠是合法的,会产生多个普通 Present。不存在单例的 Screen Overlay 泳道,也没有全局的合并步骤。
-
-## 4. 时间与局部包络
-
-Selection、Moment、Program 和绝对时间都通过 [`track-authoring.md`](./track-authoring.md) 下沉。Screen Overlay
-没有自己的 `from`、`until`、`manual` 或短语匹配式的时间法则。
-
-投影出的窗口只回答条目在哪里存在。窗口内的包络属于组件呈现。例如 Flash 可以定义 attack、峰值和 decay;
-Directional Matte 可以把一条边从一侧移到另一侧;Grain 可以保持稳定。旧的那个在每个效果时长 30% 处达到峰值的通用三角包络,作为隐藏的呈现策略被废弃。
-
-Recipe 默认值可以提供包络,但解析后的 Program 必须包含精确的参数。效果不能为了容纳动画而扩张自己投影出的窗口。作者写得更长的包络只会被该窗口裁剪;它不会被拒绝,也不会被偷偷加速。非法或零长度的窗口仍然通过公共时间 validator 失败。
-
-## 5. 可移植的官方组件集
-
-有用的旧效果按两种诚实的 lowering 方式分为两组。
-
-### 直接由 Visual IR 承载的候选
-
-它们通常可以 lower 成全画布的矩形、渐变和局部关键帧:
-
-- `Flash` —— 带显式不透明度包络的纯色;
-- `ColorWash` —— source-over 的色彩薄纱;没有依赖下层像素的隐藏混合模式;
-- `Vignette` —— 作者声明的中心、半径、柔和度、圆度、颜色和不透明度;
-- `ScanLines` —— 间距、粗细、角度、不透明度和局部运动;
-- `DirectionalMatte` —— 角度、覆盖范围、羽化、颜色、不透明度和进度;
-- `WhipVeil` —— 方向、宽度、柔和度、行程和不透明度;
-- `GlitchVeil` —— 确定性的、自有的色带 / 色块,而不是对下层像素的位移。
-
-### 自有 Surface 的候选
-
-把它们做成确定性的、带 alpha 的 `CompositableSurface` 值可能更容易或更精确,但当前的官方实现正好能完全落在封闭的、无代码的 Visual IR 里:
-
-- `Grain` —— 数量、颗粒尺寸、单色 / 彩色、运动速率和必需的 seed;
-- `LightLeak` —— 颜色、角度、柔和度、行程、强度,以及随机时必需的 seed;
-- `Bokeh` —— 数量、尺寸范围、颜色 / 色温、漂移和必需的 seed;
-- `TVStatic` —— 数量、噪点尺寸、扫描线贡献、运动速率和必需的 seed。
-
-只有当两种方式都产出完全相同的声明像素时,选择 Visual IR 还是 Surface 才是 lowerer 的实现决策。它不能取决于恰好由哪个 Runtime 执行这次 Build。做实体化的实现可以发出一次普通的 Provider Need,但该 Need 返回的是带类型的自有 Surface,而不会收到下层合成结果。
-
-一切随机行为都要求一个显式解析出的 seed。`Math.random()`、墙上时钟、GPU 噪声或依赖渲染器的熵都是非法的。
-
-## 6. 明确排除的效果
-
-以下旧名称被排除在这个无源包之外:
-
-| 旧效果 | 排除原因 | 诚实的替代方案 |
-|---|---|---|
-| `gaussian_blur` | 必须读取被模糊的像素 | 带显式输入的媒体 / surface 滤镜 |
-| `color_adjust` | 曝光 / 对比 / 饱和 / 色相会改变输入像素 | 带显式输入的媒体 / surface 色彩变换 |
-| `zoom_blur` | 真正的径向模糊需要被采样的画面 | 带显式输入的媒体 / surface 效果 |
-| screen / multiply / overlay 混合薄纱 | 结果依赖下层像素 | source-over alpha 近似,或带显式输入的实体化 |
-
-这并不删除视觉效果本身。它拒绝的是旧有的那种特权:一个号称普通的 Track 竟能改动它下方的一切。
-
-Screen Overlay 同样不能:
-
-- 改变 Program 时长或播放速率;
-- 拥有相邻媒体条目之间的切点;
-- 为两个兄弟 Track 做转场;
-- 预留最顶层的 z-index;
-- 把音效藏在视觉条目内部。
-
-被视觉遮盖的硬切仍然是两个独立事实:Media 包拥有那个切点,而 Screen Overlay 恰好覆盖了其中一些帧。作者组合出的 `Impact` Fragment 可以从同一个 Moment 产出一个 Visual Track 和一个 Audio Track,但两个输出仍然是进入 Film 的显式同级边。
-
-## 7. Stacking
-
-旧的固定 `z_index: 200` 规则被废弃。“Screen”描述的是全画布几何,不是某个魔法渲染阶段。每个 Present 都获得一个显式的绝对 stacking 意图,Composition 把它与其他所有 Present 一同展平。
-
-作者可以有意把暗角放在字幕之下、把闪光放在字幕之上,或者把颗粒放在两个图形层之间。包的 recipe 可以提供方便的 stacking 默认值,但 Film 和 Composition 绝不会按包的家族做分支,也绝不会强行把它顶到最上层。
-
-如果某个 overlay 包在不同的 stacking 位置发出多个 Present,同级 Track 可以按照扁平 Track 法则的要求在它们之间交错。
-
-## 8. 旧系统迁移审计
-
-保留:
-
-- 短促的 Moment 命中与长时段的 Program overlay;
-- Selection、Moment 和绝对时间放置;
-- 全画布的闪光、薄纱、暗角、漏光、颗粒、扫描线、散景和雪花视觉;
-- 确定性的效果专属参数与局部包络;
-- 多个效果同时存在;
-- 可选的显式自有纹理或实体化的 alpha Surface。
-
-废弃:
-
-- 单例槽位语义;
-- 固定的最顶层 z-index 和字幕之后的特殊渲染阶段;
-- 把一个中心字符串算子注册表当作系统事实;
-- 通用的 30% 峰值包络;
-- 隐藏的 `backdrop-filter` 访问;
-- 跨 Track 的 `mix-blend-mode`;
-- 把屏幕命中假装成场景转场;
-- 隐藏的泳道级进出音效;
-- 任何 Base FX 对应物或占位实现。
-
-## 9. Lowering 与验收
-
-实现遵循共享的 Temporal 合同和既有的 Visual Track 合同:
-
-1. **已实现:** 自描述的 `@narratage/screen-overlay` 作者包;
-2. **已实现:** 十一个相互独立的带类型作者组件,而不是一个不做校验的口袋;
-3. **已实现:** 精确的、无代码的 `svml.visual-ir@1` lowering;当未来某个组件无法被精确表达时,通用的带类型
- 自有 Surface 逃生口仍然可用;
-4. **已实现:** 使用连接进来的 CanvasSpace 与 ProgramSpace,绝不使用 Film 全局量;
-5. **已实现:** 普通的绝对 stack Present;
-6. **已实现:** 每个组件与每种包络的确定性结构 / 帧测试;
-7. **已实现:** 单 worker 与三 worker 的真实 Chromium 渲染产出完全相同的像素;
-8. **已实现:** 多个 Overlay Track 按 stacking 键与同级 Track 交错;
-9. **已实现:** 拒绝任何下层合成结果、兄弟 Track、backdrop-filter 或隐藏音频依赖的反例测试。
-
-如果新增一个 overlay 需要在 Core 里加分支、在 Composition 里加家族、在 HyperFrames 里加组件开关,或者需要对已合成像素做隐式快照,那么这个设计就是失败的。
diff --git a/docs/zh/guide/spec/script-surface.md b/docs/zh/guide/spec/script-surface.md
deleted file mode 100644
index c0e0fc9c..00000000
--- a/docs/zh/guide/spec/script-surface.md
+++ /dev/null
@@ -1,602 +0,0 @@
----
-title: SVML Script Surface
-description: Script 作者面的规范性合同:Segment、Role Cue、Dual Text、Selection、Moment 与 Slot。
----
-
-# SVML Script Surface
-
-> **Draft; not frozen.**
->
-> SVML 尚未公开发布,不存在需要兼容的旧语言版本。本文直接定义首个公开目标
-> Script Surface v1;仓库中的早期原型不是规范合同。实现只有在通过本文的
-> parser、projection、source-map 和 temporal golden fixtures 后,才可声称
-> 支持 SVML v1。
-
-SVML 是语义视频源语言,不是像素渲染格式。本文只定义它最高频、最需要保持可读性的稿子区 `<script>`;画布、节点、样式、生成参数和其他 Program 区域的外层文档结构另行设计。v1 发布后,新增语义应首先组合本文已有构造;确需改变正文语法时再发布新版本,不能因为早期原型行为而削弱当前 v1。
-
-## 一页合同
-
-正文除自然语言外只有六种语义构造:
-
-| 构造 | 只负责 | 不负责 |
-|---|---|---|
-| `Segment` | 有序的媒体/口播块和独立结构端点 | 自动 Selection、时长、静音 |
-| `Role Cue` | `dialogue` 导出时的“谁说了什么”文本前缀 | speaker 实体、音色、字段路由 |
-| `Dual Text` | 显示文字与实际读音不同 | TTS 厂商参数、样式、动作 |
-| `Selection` | 一个或多个显式闭合的语义时间区间 | 消费者声明、绝对秒数 |
-| `Moment` | 一个或多个显式完整的语义时刻 | 半截区间、持续时间 |
-| `Slot` | 注入运行时文字值 | 注入 SVML 语法或任意 AST |
-
-注释是 trivia,不是第七种语义构造。任何字幕样式、B-roll、Ranking、Deck、音效、静音、生成提示、节点端口和消费策略都在 `<script>` 外引用
-Selection 或 Moment。消费者不归属于某个 Segment;一个消费者可有多个端口,每个端口可接一个或多个 Selection/Moment。Selection 可以跨 Segment
-或由非连通区间组成,Moment 也可以有多个 occurrence。
-
-Script 编译必须先生成保留源码映射的 Narrative IR,再从同一份 IR 投影三种文本。禁止先生成三个互不相关的字符串,再猜它们之间的对应关系。
-
-| 投影 | 内容 | 典型消费者 |
-|---|---|---|
-| `dialogue` | Role Cue + 实际读音 | Seedance 等对话/视频生成 |
-| `speech` | 仅实际读音 | Estimate、TTS、Speech Reference、对齐 |
-| `caption` | 仅显示文字 | 字幕与文本展示 |
-
-## 正例
-
-```svml
-<script>
- @whole
-
- <hook>
- <A> I just @laugh @pop! <lmao | laughed my @punch ass out @/punch> @/laugh.
- </hook>
-
- @silence
- <pause/>
- @/silence
-
- <close>
- <B> Meet <${product} | ${product_pronunciation}>.
- </close>
-
- @/whole~
-</script>
-```
-
-若 `product = "SVML"`、`product_pronunciation = "ess vee em ell"`,三个文本投影分别是:
-
-```text
-dialogue
-A: I just laughed my ass out.
-B: Meet ess vee em ell.
-
-speech
-I just laughed my ass out.
-Meet ess vee em ell.
-
-caption
-I just lmao.
-Meet SVML.
-```
-
-`silence` 只是 Selection 的名字,不会自动让音频静音。它完整选择了空
-Segment 的两个独立结构端点;真正的静音或素材行为由外部 Program 决定。
-`pop` 是一个默认吸右的 Moment,解析到 `laughed` 的起音;它同样不进入任何文本投影。
-
-## 1. 文档与 Segment
-
-在官方 Markup Frontend 中,一个 `<script>` Surface 实例产生一份 Narrative;
-`<script>` 外壳由 Text 根据导入后的 Surface Registry 分派,Script Parser 接收的只是它的 raw body。Script body 的 Segment 形式只有具名块和具名空块:
-
-```svml
-<intro>
- Hello.
-</intro>
-
-<pause/>
-```
-
-规范规则:
-
-- Script 包含一个或多个按源码顺序排列的 Segment。
-- Segment 的标签名就是 id;id 在当前 Script 内唯一,Segment 不允许嵌套。
-- Segment opening 不接受 attribute;`script` 是外层 raw Surface 的保留名,不能
- 作为 Segment id。
-- 自然语言只能出现在 Segment 内;Segment 之间只允许空白、Selection/Moment
- 标记和注释。
-- 空 Segment 使用自闭合形式。它有结构开始/结束切点,但没有词法锚点。
-- 空 Segment 不蕴含静音,也不蕴含任何默认时长。其物理时长来自媒体、生成
- 结果或 `<script>` 外的 Program。
-- Segment 不自动声明同名 Selection。选择一个、三个或任意多个 Segment,
- 都必须显式写闭合 Selection。
-- 每个 Segment 独立拥有 `start` 与 `end` 两个稳定 identity;相邻 Segment
- 不共享结构端点 identity。两个 identity 可以在硬切时落到同一 ProgramPoint,
- 也可以因 overlap 或 gap 落到不同点,但不能合并身份。
-
-例如选择三个连续 Segment:
-
-```svml
-@chapter
-<one>
- One.
-</one>
-<pause/>
-<two>
- Two.
-</two>
-@/chapter
-```
-
-## 2. Role Cue
-
-Role Cue 只在 Parser 已进入某个 Segment body 后出现,并开启一个 spoken turn。它的识别由解析状态决定,与逻辑行首和物理换行无关:
-
-```svml
-<dialogue>
- <A> What time
- is it?
- <B> It’s 8:30.
-</dialogue>
-```
-
-下列紧凑写法具有完全相同的 Narrative 语义:
-
-```svml
-<dialogue><A> What time is it?<B> It’s 8:30.</dialogue>
-```
-
-它在 `dialogue` 投影中由规范 serializer 输出为:
-
-```text
-A: What time is it?
-B: It’s 8:30.
-```
-
-在 `speech` 和 `caption` 投影中,`<A>`、`<B>` 都被移除。Role Cue 不是
-speaker 数据模型,不建立人物实体,不选择音色,也不对字幕隐式分组。外部
-Program 可以显式写 `role="A"` 查询这些 spoken turn。Caption Program 将这类查询直接降低为 `CaptionDisplayWordSubset`;它不是时间 Selection,也不伪造 Selection
-Record。仅有 `<A>` 本身不触发任何样式、人物、音色或素材行为。
-
-为消除歧义:
-
-- 在 Segment body 状态中,不含未转义 `|`、没有属性或 `/`、且满足 Role label
- 字符约束的裸尖括号构造被识别为 Role Cue;它必须在下一个 Role Cue 或 Segment
- 结束前跟随非空 spoken content。
-- 正文里的冒号永远是正文;解析器不通过 `A:` 猜说话人。
-- 同一个非空 Segment 要么完全无 Role Cue,要么第一个 spoken atom 必须由
- Role Cue 开启。时间标记、注释和布局空白不算 spoken atom;已经以无 Cue
- 正文开始的 Segment 不得在中途切换为有 Cue 模式。
-- 一个 Role Cue 的 turn 一直延续到下一个 Role Cue 或 Segment 结束。turn
- 内的物理换行只是布局空白,不结束 turn,也不要求重复 Role Cue。
-- Role Cue 不是容器标签,没有 close syntax;`</A>`、`</B>` 等形式必须作为
- 未知尖括号构造失败,formatter 永不输出它们。
-- Role label 是 NFC 后 1–32 个 Unicode 字符的可读文本;可由 Unicode
- Letter、Mark、Number、内部空格、`_`、`-`、`.` 组成,首尾不得有空白。
- 引号、`=`、`/`、换行、`<`、`>`、`|`、`:` 均非法,因此带属性的未知
- 结构标签不会被吞成 Role Cue。
-- 在 `dialogue` 投影中,Role Cue 边界由规范 serializer 输出为一次换行和
- `label: ` 前缀;在 `speech` 与 `caption` 中只移除 Cue,并在相邻 atom
- 之间保留必要分隔。重新折行源码不得改变 Narrative IR 的语义内容或任一
- 文本投影;允许变化的只有 source range 和 source hash。
-
-## 3. Dual Text
-
-Dual Text 同时写出显示文字和实际读音:
-
-```svml
-<lmao | laughed my ass out>
-```
-
-左侧进入 `caption`,右侧进入 `dialogue` 和 `speech`。普通正文是
-`<text | text>` 的简写,不需要任何标记。
-
-Dual Text 是一个不可嵌套的原子对应单元,而不是简单的字符串替换:
-
-- 两侧可以是任意 N:M 文本映射,不要求词数相等。
-- 右侧去除布局空白后必须非空;只显示不发声的文字不属于 Script,应由外部
- Text/Deck/Caption Program 表达。
-- 左侧可以为空,例如 `< | um>` 表示实际说出 filler,但字幕不显示。
-- Slot 两侧都可使用;Selection/Moment 只可出现在 speech 侧,并且只能落在
- 其词法 token 边界。
-- 整个显示侧继承整个 speech 侧的时间跨度。若显示词需要独立计时、选择或
- 样式,作者必须把它拆成多个 Dual Text 原子。
-- 普通时间消费者可以选择一个 Dual Text speech 侧的部分 token;但字幕
- ownership/style 消费者若只覆盖一个 Dual Text 原子的部分 speech span,
- 编译必须以 `partial dual atom` 失败,不能猜测如何切左侧显示文字。
-
-例如合法的细粒度时间选择:
-
-```svml
-<lmao | laughed my @middle ass out @/middle>
-```
-
-`middle` 可以供 B-roll 或音效使用;若它被用作该 `lmao` 的局部字幕替换区,则必须报错。作者可通过拆分原子提供明确映射。
-
-Dual Text 只表达“实际说了什么”。IPA、SSML、重音、语速、语种、情绪、音色和厂商专用发音字典不是第四种正文文本层,应由外部生成/语音 Program
-引用 Selection 或词典资源表达。
-
-## 4. Selection
-
-Selection 是由作者显式闭合的语义区间声明:
-
-```svml
-@id ... @/id
-~@id ... @/id~
-```
-
-每个开始标记都必须有同 id 的结束标记。Script 的公开值为每个 occurrence
-输出已经解析好的 `startAnchorId` 与 `endAnchorId`;不存在 start-only 对象、消费者默认补尾或“信息传了一半”的状态。左右 affinity、token 下标和源码范围属于 Parser/Source Map,不泄漏到公开 Selection。需要物理时间的下游必须把这两个身份与显式连接的 `CompleteSemanticMap` 投影为区间。
-
-“闭合”指源码的两个端点都被明确声明。物理运行时使用半开
-`[startFrame, endFrameExclusive)` Frame Span。
-
-### 4.1 左右吸附
-
-标记本身零宽,不进入任何文本投影。它位于左右两个可选语义边界之间:
-
-| 端点写法 | 吸附方向 | 典型含义 |
-|---|---|---|
-| `@id` | 右 | 从右侧词首或右侧结构端点开始 |
-| `~@id` | 左 | 向外扩到左侧词尾或左侧结构端点 |
-| `@/id` | 左 | 在左侧词尾或左侧结构端点结束 |
-| `@/id~` | 右 | 向外扩到右侧词首或右侧结构端点 |
-
-例子:
-
-```svml
-before @x hello @/x after
-```
-
-`x` 默认从 `hello` 的词首到 `hello` 的词尾。
-
-```svml
-before ~@x hello @/x~ after
-```
-
-`x` 向两侧外扩,分别取 `~@x` 左侧和 `@/x~` 右侧最近的可选边界。
-
-```svml
-@pause
-<empty/>
-@/pause
-```
-
-`pause` 的开始吸附空 Segment 的开始切点,结束吸附它的结束切点。Selection
-可完全位于 Segment 外,也可跨任意数量的 Segment。
-
-### 4.2 闭合、交叉与非连通
-
-同一个 id 可以顺序出现多次,编译为一个含多个独立锚点对的非连通 Selection:
-
-```svml
-@beat one @/beat ... @beat three @/beat
-```
-
-不同 id 可以自由交叉,不要求 XML 式嵌套:
-
-```svml
-@a one @b two @/a three @/b
-```
-
-规范要求解析器按 id 管理开放状态,而不是使用一条全局栈。以下情况必须失败:
-
-- 未闭合、孤立 close 或 open/close id 不一致;
-- 同一个 id 尚未闭合就再次 open;
-- Selection 标记切进一个 v1 speech token 内;
-- 依赖消费者默认 span 来补齐缺失端点。
-
-同一 id 的多个 occurrence 保持源码顺序和稳定 occurrence identity。不同
-SelectionSet 以及同一集合的不同 occurrence 可以在物理时间上重叠。
-
-### 4.3 可选时间边界
-
-Slot 绑定、Dual Text speech 投影和 NFC 归一化之后,v1 使用规范定义的确定性 `speech-tokenizer-v1`:
-
-- 东亚表意文字和假名单字符成 token;
-- 其他受支持文字/数字脚本形成最大连续 run;
-- `'` 或 `’` 只在两个 run 字符之间时保留;
-- 空白、标点、emoji 和其他分隔字符不产生词法 token。
-
-因此标点本身不可被计时,Selection/Moment 也不可切入单个 token。v1
-编译器在移除零宽标记后再 tokenization,所以未来若需要字素、音节或音素边界,可以在新的 timing profile / Script Surface 版本中复用相同的
-Selection/Moment 类型;v1 不应为尚不存在的精度引入新正文符号。
-
-若 Script 有 `N` 个 Segment,第 `k` 个 Segment 有 `mₖ` 个 speech token,且
-`M = Σmₖ`,Semantic Anchor Index 恰有:
-
-```text
-Σ(2mₖ + 2) = 2M + 2N
-```
-
-个稳定身份。每个 Segment 内的局部顺序是:
-
-```text
-segment[k].start
-token[k,1].start
-token[k,1].end
-...
-token[k,mₖ].start
-token[k,mₖ].end
-segment[k].end
-```
-
-Program 起点和终点属于 ProgramBasis,不额外进入 Semantic Anchor Index。空
-Segment 仍有独立 start/end,即使二者最终重合;任意不同 identity 即使最终落到同一帧也不能合并。
-
-Locator 必须提交覆盖全部 `2M + 2N` identity 的总映射:
-
-```text
-SemanticAnchorIdentity → ProgramPoint(ProgramSpace)
-```
-
-标记的左右吸附在 Script 解析时就确定到唯一一个 anchor。Token 切点与 Segment
-切点在此完全平等:段首标记左吸附取该 Segment 自己的起始切点,绝不越到前一个
-Segment 的词尾。下游拿到的是 anchor 身份,不再自行做下标推算。
-
-定位是全覆盖的:每个 Segment 的每个 token 都带有窗口,无论它被测得、由相邻字符推得,还是因为转写从未触及而由插值补出。窗口倒序、越出所属 Segment 或与邻居重叠,都按测得原样报出;那是关于录音的事实,如何解释属于把它投影到时间轴的那一方。定位只在 Script、音频与转写不是同一组三件事时失败,绝不因为一个时间戳而失败。
-
-每个点都必须存在。消费者不得补点、移动点或从相邻 occurrence 借点。每个 Segment
-内部必须非降序:
-
-```text
-segment.start ≤ token₁.start ≤ token₁.end ≤ ... ≤ segment.end
-```
-
-按源码相邻的 Segment `A`、`B` 还必须保持:
-
-```text
-A.start ≤ B.start
-A.end ≤ B.end
-```
-
-这允许硬切、重叠与留白,同时禁止一个 Locator 把后写的整个 Segment 静默排到前写 Segment 之前:
-
-```text
-hard cut A.end == B.start
-overlap B.start < A.end
-gap B.start > A.end
-```
-
-若未来需要真正重排或并行 speech,应发布显式的非线性叙事模型;v1 不让普通
-Range 在不同 Locator 下反向或消失。
-
-同一份 Selection 在不同 fulfillment 产生的 `CompleteSemanticMap` 上可得到预览或成片区间。Map 使用相同 identity。Script 本身不含秒数、帧号或采样点。
-Script Surface 不绑定帧率、采样率或渲染器。后端一旦选择物理时钟,必须只量化一次并让所有消费者复用同一整数边界;后端时钟变化不改变本语言表面。
-
-对齐、结构端点、projection 或帧量化可能使某个 occurrence 最终成为零长或反向区间。实现必须产生明确诊断,不能静默丢弃、移动 Selection 端点、借用相邻 occurrence 或替作者补范围。
-
-## 5. Moment
-
-Moment 是完整的点声明,不是省略了 close 的 Selection:
-
-```svml
-@id!
-~@id!
-```
-
-`!` 明确标出这个名字的类型是 Moment。它不需要也不允许 close:
-
-| 写法 | 吸附方向 | 典型含义 |
-|---|---|---|
-| `@id!` | 右 | 右侧词首或右侧结构端点 |
-| `~@id!` | 左 | 左侧词尾或左侧结构端点 |
-
-默认右吸覆盖“与某个词一起开始”的高频情况:
-
-```svml
-I just @pop! laughed my ass out.
-```
-
-`pop` 解析到 `laughed` 的起音。显式左吸写成:
-
-```svml
-I laughed ~@pop! and left.
-```
-
-`pop` 解析到 `laughed` 的收音。位于两个 Segment 之间时,左候选是前一
-Segment 的 `end`,右候选是后一 Segment 的 `start`;两者可以同点,也可以因
-overlap 或 gap 落在不同点。
-
-规范规则:
-
-- 同 id 的 `@id!` / `~@id!` 可以重复,按源码顺序编译成一个 Moment;每个
- occurrence 的公开值只有一个已解析的 `anchorId`。
-- Selection 与 Moment 共享 temporal name namespace。同一个 id 不能同时
- 声明两种类型;`@x! ... @/x` 必须以类型冲突失败。
-- 单独的 `@x` 永远是未闭合 Selection open,必须报错;解析器不得因为没有
- 找到 `@/x` 而把它猜成 Moment。
-- `@id!~` 是冗余且非法的右吸写法;右吸只写规范形式 `@id!`。
-- 语法只接受 ASCII `!`。未转义 `@id!`(全角感叹号)是 malformed
- temporal syntax,不能退化为正文。
-- Moment 与 Selection marker 可出现在相同位置,包括 Segment 之间和
- Dual Text 的 speech 侧,但均不得切入 v1 speech token。
-- Moment 只选择 Semantic Anchor Index 中已有的一个候选点,不新增 anchor
- identity;因此 `2M + 2N` 的精确计数不因 Moment 数量改变。
-- Moment 与 Selection 复用同一份 CompleteSemanticMap 和一次性
- 帧量化;区别只在最终载体是 `Point[]` 而不是 `Range[]`。
-
-消费端保持两个互不相混的类型:
-
-```text
-NarrativeSelection = { startAnchorId, endAnchorId }[]
-NarrativeMoment = { anchorId }[]
-```
-
-推荐的显式使用关系是:
-
-```svml
-during="phrase" <!-- 只接受 SelectionSet -->
-at="pop" <!-- 只接受 MomentSet -->
-at="phrase.start" <!-- SelectionSet 的各 range 起点投影 -->
-at="phrase.end" <!-- SelectionSet 的各 range 终点投影 -->
-```
-
-`at="phrase"` 不得默认猜成 start,`during="pop"` 也必须类型报错。一个消费者若同时需要窗口和触发点,必须使用两个字段/端口,而不是构造
-`Range | Point | mixed[]` 的多态 TemporalRef。端口需要 `one` 还是 `each`
-属于消费者 cardinality 契约,不改变 Script 的单/多 occurrence 语法。
-
-## 6. Slot
-
-当前实现状态:`parseScript` 已实现下述 parse-first、literal-only 绑定合同,但官方
-`<script>` Author Surface 尚未暴露 bindings 输入。未绑定 Slot 在当前 `main.svml` 中会按
-`SCRIPT_SLOT_UNBOUND` 失败;公开使用示例在 Surface 补齐绑定边之前不得宣称 Slot 可直接使用。
-
-Slot 是解析器识别的一等 AST atom:
-
-```svml
-${product}
-<${product} | ${product_pronunciation}>
-```
-
-Slot id 经 NFC 后必须由 1–64 个字符组成:首字符为 Unicode Letter 或 `_`,其余字符可为 Unicode Letter、Mark、Decimal Number、`_` 或 `-`。绑定值是单行 Unicode 纯文本,不能含 CR/LF 或控制字符。
-
-关键安全规则:
-
-- 先解析 Script,再绑定 Slot;禁止在解析前做字符串替换。
-- 绑定值永远按 literal text 进入已经存在的 atom,不能注入 Segment、Role
- Cue、Dual Text、Selection、Moment、Slot、注释或转义序列。
-- 未绑定 Slot 可以保留在结构化编辑 IR 中,但凡是生成三种文本、计算 token
- 或编译时间线,都必须 fail closed。
-- SVML v1 writer 只输出 `${变量}`;未转义的 `[变量]` 只是普通正文。
-
-## 7. 词法保留、转义与规范排版
-
-源码统一把 CRLF/CR 转成 LF,并做 Unicode NFC。源语言中未转义的 `<`、`@`
-和 `${` 是保留语法起点。格式错误的保留语法必须报错,不能退化为 spoken
-text。
-
-正文转义:
-
-```svml
-\@openai
-\<not-a-role>
-\${literal}
-\\
-```
-
-Dual Text 内另有:
-
-```svml
-\|
-\>
-```
-
-未知反斜杠转义必须报错。解析优先级固定为:
-
-1. Script body 状态中的注释、temporal marker 与具名 Segment open;
-2. Segment body 状态中的当前 Segment close;
-3. Segment body 中含一个未转义分隔 `|` 的 inline `<display | speech>`;
-4. Segment body 中满足 Role label 合同的裸 `<Role>`;
-5. 其他尖括号结构报错。
-
-Segment 与 temporal name 使用独立命名空间;Selection 和 Moment 共享
-temporal name namespace。三者的 id 均满足:
-
-```text
-[a-z][a-z0-9_-]{0,63}
-```
-
-源码缩进、换行和连续空白全部是 authoring layout,不参与 Segment、Role Cue
-或 turn 的识别,也不是字幕换行或停顿指令。编译投影在 atom
-边界保留必要分隔并规范化布局空白;字幕 cue 切分与硬换行由外部 Caption
-Program 决定。
-
-结构标签和注释不依靠独占一行才被解析。规范 formatter 把结构标签和注释各自放在独立行,使用两空格缩进 Segment、四空格缩进 Segment 内的 spoken
-content,并在 Segment 之间留一空行。formatter 前后重新解析所得 Narrative
-IR 的语义内容必须相同;允许变化的只有 source range 和 source hash。
-
-注释使用 `<!-- ... -->`,可以跨行但不能嵌套,也不能写进 Dual Text、Slot、
-temporal marker、Role Cue 或结构标签内部。注释可出现在 atom 之间允许布局空白的位置,并按布局空白处理;规范 formatter 将其独占一行。注释不进入任何投影、token、Selection、Moment 或 hash 的语义内容;需要可复现源码身份时可以另算 source hash。
-
-## 8. Narrative IR 与消费者合同
-
-解析结果至少保留:
-
-- Segment 的顺序、id 和独立 start/end identity;
-- 每个 Segment 的有序 spoken turn,以及各 turn 的稳定 identity、可选 Role Cue、
- token range 和 source range;
-- plain / Dual Text / Slot atom 及其源码范围;
-- Caption display Atom、其内部有序 display Word,以及 Atom 到一个或多个 speech token
- 的独立显式映射;
-- 每个 Selection 的一个或多个有序 occurrence,以及各 occurrence 已解析的
- `startAnchorId` / `endAnchorId`;
-- 每个 Moment 的一个或多个有序 occurrence,以及各 occurrence 已解析的
- `anchorId`;
-- Parser/Source Map 内部另行保留 affinity、token/source range,公开值不携带它们;
-- 完整有序的 `CaptionDisplaySequence`、独立的 `CaptionCorrespondence`,以及每个显式
- Selection 对应的 `CaptionDisplayWordSubset`。字幕无需也不得从公开 Selection 反推词下标。
-
-外部 Program 的 `role="label"` selector 可以从上述词全集投影自己的词子集;同一 label 的多次 Turn 是多个有序、互不相邻的命中。它是 Caption author surface
-内部的显式查询,不是 Role Cue 的隐式字幕行为,也不建立 speaker entity。
-
-SelectionSet/MomentSet 是消费者边界,不是 Segment 的子对象。典型外部关系是:
-
-```text
-Script ──compile──> Narrative IR ──project──> dialogue / speech
- ├──resolve──> NarrativeSelection[] / NarrativeMoment[]
- └──project──> CaptionDisplaySequence
- + CaptionCorrespondence
- + CaptionDisplayWordSubset[]
-
-NarrativeSelection A ──during──> consumer.port_1
-NarrativeMoment P ──at──────> consumer.port_2
-CaptionDisplayWordSubset C ──words──> caption.Program
-```
-
-一个 Ranking、B-roll 或其他节点可以有多个动态端口,各端口接受不同
-Selection/Moment;同一集合也可以被多个消费者复用。未来 Graph
-语法可以把这些类型边写成 `during=` / `at=`,但它们必须引用已解析的名字或端点投影,不能接受模糊自然语言 locator。
-
-Caption 的多样式、region、cue segmentation、annotation、mute 和 layout
-都属于 Caption Program。Caption consumer 若要求互斥 token ownership,重叠必须作为该消费者的编译错误;这不是 Script 禁止 Selection 重叠。未来 Caption 若支持 overlay,只扩展 Caption Program,不扩展正文语法。
-
-当前 Caption Program 使用一个覆盖全部可见词的默认 Style,再按源码顺序应用 Role
-或显式 Selection 的整 Style 替换,最后命中的规则获胜。无 Role 的 Turn 自动保留默认
-Style;修改一个局部区间不要求作者书写它的补集。完整规则见
-[`caption-program.md`](./caption-program.md)。
-
-## 9. v1 明确不在 Script 中表达
-
-以下能力不是遗漏的第七种 inline marker:
-
-| 需求 | v1 归属 |
-|---|---|
-| 绝对秒数、帧号、采样点、任意 sub-frame 位置 | Script 外的运行时/Program |
-| token 内音素或字素级切点、标点专属时间 | 后续 timing profile 或新版本 |
-| 空 Segment 的预览时长 | 媒体、生成器或显式 Program |
-| 同时重叠说话 | 多轨媒体/生成 Program;Base Script 仍是顺序语义轴 |
-| 笑声、咳嗽、环境声、SFX、beep | 音频/生成/SFX Program |
-| 无 speech 时间依据的纯显示文字 | Text、Deck 或 Caption Program |
-| 字幕样式、手工断行、动效、画面动作 | 相应 Program 引用 Selection |
-| 语气、情绪、音色、重音、语速、SSML | Speech/Generation Program |
-| Consumer、节点、端口、DAG 接线 | `<script>` 外的 SVML 区域 |
-
-这些边界保证 Script 保持“可直接阅读的稿子”,同时不阻止完整 SVML 最终确定性编译 DAG 和时间线。
-
-## 10. 完备性与扩展判定
-
-遇到新需求时按下列顺序归类:
-
-1. 显示文字和实际读音不同:Dual Text。
-2. 对话导出的朗读人前缀不同:Role Cue。
-3. 语义时间范围不同:Selection。
-4. 语义瞬间不同:Moment。
-5. 顺序媒体/口播块不同:Segment。
-6. 运行时文字输入不同:Slot。
-7. 字幕、画面、音频、生成或消费行为不同:Script 外的 Program。
-
-只要需求能落入这七项,就不得发明新的 inline delimiter。若七项都不能无损表达,应先证明它属于稿子语义而不是 Program,再以新版本提案处理;不得让现有解析器“顺便支持”。
-
-## 11. 实现验收
-
-一个实现只有同时满足以下证据,才能声明支持 Script Surface v1:
-
-1. 本文全部正例、反例、投影和 source-map golden fixtures;
-2. parser、formatter、Narrative IR 与三投影不依赖 provider 或渲染器;
-3. SelectionSet 支持闭合、交叉、非连通 occurrence 与左右 affinity;
-4. MomentSet 支持 `@id!` / `~@id!`、单/多 occurrence 与 `at` 类型闸;
-5. Dual Text 在 speech、caption 与时间映射上保持一个可审计 source map;
-6. Slot 采用 parse-first、literal-only binding,不能注入语法;
-7. 未知版本、未知保留语法、混合 temporal type 与 partial dual atom 全部
- fail closed。
-8. `2M + 2N` identity 的规范序列化、digest、完整 Map 与相邻 Segment 双单调
- 约束都有 golden fixtures;硬切、overlap、gap、空 Segment 和端点重合均被
- 覆盖。
diff --git a/docs/zh/guide/spec/spatial-layout.md b/docs/zh/guide/spec/spatial-layout.md
deleted file mode 100644
index 3a4d59fe..00000000
--- a/docs/zh/guide/spec/spatial-layout.md
+++ /dev/null
@@ -1,406 +0,0 @@
----
-title: 共享 Spatial 布局
-description: 共享 Spatial 几何的可执行基础权威。`@narratage/spatial`。
----
-
-# 共享 Spatial 布局
-
-状态:共享 Spatial 几何的可执行基础权威。`@narratage/spatial` 实现了 CanvasSpace、Frame/Point/Path、
-IntrinsicExtent、ContentFit 以及纯粹的 fitting;Film 现在消费与 Track 布局完全相同的显式 CanvasSpace。
-Text、Media 以及四个 Ranking witness 均可执行,仓库内部的 Visual IR 兼容性审查已通过。公开的包 ABI
-发布仍然是另一件事。
-
-## 1. 结论
-
-共享 spatial 系统拥有的是坐标几何,而不是 Text、Media 或 CSS。它的稳定词汇是:
-
-```text
-CanvasSpace exact output coordinate system
-SpatialPoint one point in that system
-SpatialFrame one positive rectangular region
-SpatialPath one explicit vector path
-IntrinsicExtent one content item's natural two-dimensional size/aspect
-ContentFit how an extent is sized and aligned against a Frame
-FittedContent the deterministic result of that fit
-```
-
-当一个可视组件消费某个 `SpatialFrame` 时,该 frame 扮演的角色是它的 **Placement Frame**:组件被摆放在哪里,以及 frame 级别的 Paint 可以存在于哪里。随后每一个实际的 text/media 样本都会依据自身的固有尺寸和一个 `ContentFit` 推导出各自的 **Content Frame**。
-
-这就是那两个 frame。它们并不总是嵌套关系:
-
-```text
-contain cover
-
-Placement Frame Content Frame
-┌──────────────┐ ┌──────────────────┐
-│ ┌────────┐ │ │ ┌────────────┐ │
-│ │Content │ │ │ │ Placement │ │
-│ │ Frame │ │ │ │ Frame │ │
-│ └────────┘ │ │ └────────────┘ │
-└──────────────┘ └──────────────────┘
-```
-
-因此把第二个称作“内框”是错误的。在 `cover`、缩放或局部运动之下,Content Frame 可能超出 Placement
-Frame。
-
-一个 Placement Frame 可以包含若干个显式样本。纯色底衬没有 Content Frame。一个模糊样本和一个前景样本各自拥有独立的 Content Frame,即使它们引用同一个 Artifact。不存在被所有图层共享的那种全局唯一“内框”。
-
-## 2. 边界
-
-聚焦的视频领域包是 `@narratage/spatial`。它拥有纯类型、validator、面向可复用几何的作者 Surface 以及确定性的几何 Producer。它不拥有:
-
-- 时间窗口或媒体的固有播放行为;
-- 文字折行或字形测量;
-- 媒体解码、流选择或容器旋转;
-- 背景、模糊、不透明度、颜色、描边或阴影;
-- z-order、转场或局部运动;
-- Composition、Runtime、Provider 或 Core 的行为。
-
-这些事实只通过图的边与 Spatial 相遇。例如,Media Inspection 选定一路可视流并产出其面向显示的 extent;某个 Media 包再显式地把该 extent 转换成 `IntrinsicExtent`,交给 Spatial 做 fitting。Spatial 永远不会打开一个 MP4,也不会猜测哪一路流是主流。
-
-安装或修订这个包不得改变 Core。Core 看到的只是普通的带类型 Record 和 Operation。
-
-## 3. CanvasSpace
-
-`CanvasSpace` 在任何 Track 被撰写之前就确立了精确的坐标系:
-
-```ts
-type CanvasSpace = {
- readonly contract: "svml.canvas-space@1";
- readonly widthPx: number;
- readonly heightPx: number;
- readonly origin: "top-left";
- readonly xDirection: "right";
- readonly yDirection: "down";
- readonly pixelAspect: "square";
-};
-```
-
-宽和高是正整数。颜色、clear Paint、帧率和时长不是空间事实,都留在这个值之外。
-
-Canvas 必须先于 Track 声明,随后作为同一个图上的值连接到 Film/Composition。当前的 Film Surface 把宽高藏在最终的 Film 声明内部,这对精确的 Text 和 Media 布局来说太晚了。它的重写必须把 Canvas 声明与最终的 Track
-装配分开,而不是把宽高复制进每一份 Track Recipe。
-
-预期的源码形态是:
-
-```svml
-<space:Canvas id="vertical" width="1080" height="1920"/>
-
-<!-- Track declarations consume vertical or Frames derived from it. -->
-
-<film:Film id="main" canvas={vertical} space={program} appearance={studio.film.main}>
- ...
-</film:Film>
-```
-
-精确的 Surface 写法仍处于冻结前状态。单值数据流是规范性的。
-
-## 4. SpatialFrame 的构造
-
-一个已解析的 `SpatialFrame` 是 Canvas 像素坐标系中的有限矩形:
-
-```ts
-type SpatialFrame = {
- readonly contract: "svml.spatial-frame@1";
- readonly xPx: number;
- readonly yPx: number;
- readonly widthPx: number;
- readonly heightPx: number;
-};
-```
-
-宽和高必须为正。`xPx` 与 `yPx` 可以为负,也可以延伸到 Canvas 之外。画布外的 Frame 对进出场运动很有用,不得被悄悄 clamp、求交或拒绝。哪些像素可见由 Canvas 光栅化决定。Clamp 永远是一次显式的投影选择。
-
-作者输入可以使用相对 Canvas 的百分比或像素。它们会在消费者布局其内容之前编译成这一种已解析形式。百分比相对的是显式连接的父 Frame 的宽/高,而不是渲染器碰巧创建的某个 DOM 容器。
-
-三个作者组件在不需要 mode 字段的前提下覆盖了常见的构造语义:
-
-1. `Frame`:在父 Canvas/Frame 内部的四条显式边;
-2. `AnchoredFrame`:一个目标点、一组宽/高,以及九个自身锚点之一;
-3. `AspectFrame`:一个目标点、一个主尺寸,以及一个显式给出或连接进来的宽高比。
-
-概念上:
-
-```svml
-<space:Frame id="safe" within={vertical}
- left="6%" top="4%" right="94%" bottom="96%"/>
-
-<space:AnchoredFrame id="card" within={safe}
- x="50%" y="78%" width="82%" height="28%" anchor="center"/>
-
-<space:AspectFrame id="sticker" within={safe}
- x="100%" y="100%" width="32%" aspect={portrait.extent} anchor="bottom-right"/>
-```
-
-`Frame` 不代表一个可见的盒子。它只产出几何。`AspectFrame` 可以消费一条通用的 `IntrinsicExtent` 边;它不依赖 Media。
-
-九个规范锚点名是:
-
-```text
-top-left top-center top-right
-middle-left center middle-right
-bottom-left bottom-center bottom-right
-```
-
-编译后的数学是点吸附,而不是九套硬编码的布局算法:
-
-```text
-frame.x = target.x - frame.width × selfAnchor.x + offset.x
-frame.y = target.y - frame.height × selfAnchor.y + offset.y
-```
-
-每个具名锚点不过是一对可读的 `(0 | 0.5 | 1, 0 | 0.5 | 1)`。更高级的 Surface 可以暴露任意归一化锚点,而不必改变编译后的合同。
-
-安全区、左栏、标题区和商品区都是从 Canvas 或父 Frame 派生出来的普通具名 Frame。它们不是保留的全局 token。
-
-## 5. IntrinsicExtent
-
-`IntrinsicExtent` 是保持内容形状所需的最小通用事实:
-
-```ts
-type IntrinsicExtent = {
- readonly contract: "svml.intrinsic-extent@1";
- readonly widthPx: number;
- readonly heightPx: number;
-};
-```
-
-这两个值是内容面向显示的、为正的自然像素 extent。`contain`、`cover`、`fit-width`、`fit-height` 和
-`stretch` 只需要它们的比值;`native` 和 `scale-down` 还会用到它们与 Canvas 像素的绝对关系。Media 可以在应用容器旋转和 sample aspect 信息之后推导它们。Text 可以从精确的字体排版推导它们。矢量组件必须选择一个显式的标称光栅 extent,而不是假装任意的 view-box 单位就是 Canvas 像素。它们都不会把源 Artifact、字体、组件族或上游摘要存进 `IntrinsicExtent`;这些关系仍然显现在图的边与 Derivation 上。
-
-## 6. ContentFit 与两个 frame 的方程
-
-`ContentFit` 选择尺寸策略、对齐方式,以及一个可选的对齐约束:
-
-```ts
-type ContentFit = {
- readonly contract: "svml.content-fit@1";
- readonly sizing:
- | "contain"
- | "cover"
- | "fit-width"
- | "fit-height"
- | "native"
- | "scale-down"
- | "stretch";
- readonly framePoint: { readonly x: number; readonly y: number };
- readonly contentPoint: { readonly x: number; readonly y: number };
- readonly offsetPx: { readonly x: number; readonly y: number };
- readonly constraint: "bounded" | "free";
-};
-```
-
-两个点都在各自的 Frame 内归一化。`framePoint` 说明所选的内容点想落在 Placement Frame 的哪个位置。
-`contentPoint` 说明源上的哪个固有点是重要的。这比旧的单一 `focalX/focalY` 更有表达力 —— 后者在两边悄悄用了同一个坐标。两个点的坐标都必须落在 `[0,1]`;有意超出该范围的位移使用 `offsetPx`,当结果必须留在范围之外时使用 `constraint="free"`。
-
-给定 Placement Frame `(Vx, Vy, Vw, Vh)` 和固有 extent `(Sw, Sh)`,保持宽高比的基础缩放为:
-
-```text
-contain min(Vw / Sw, Vh / Sh)
-cover max(Vw / Sw, Vh / Sh)
-fit-width Vw / Sw
-fit-height Vh / Sh
-native 1
-scale-down min(1, contain-scale)
-```
-
-对这些模式:
-
-```text
-Cw = Sw × scale
-Ch = Sh × scale
-```
-
-`stretch` 有意放弃源宽高比,产出 `(Cw, Ch) = (Vw, Vh)`。它之所以是显式的,是因为成熟的编辑器允许形变;其他任何尺寸模式都不得形变。
-
-约束之前:
-
-```text
-Cx = Vx + Vw × framePoint.x - Cw × contentPoint.x + offset.x
-Cy = Vy + Vh × framePoint.y - Ch × contentPoint.y + offset.y
-```
-
-对 `free` 而言,这些坐标就是最终结果。对 `bounded` 而言,每条轴按同一条法则做最小限度的 clamp:
-
-- 内容较小时,让完整的 Content Frame 留在 Placement Frame 内部;
-- 内容较大时,让完整的 Placement Frame 被 Content Frame 覆盖。
-
-对水平轴来说这意味着:
-
-```text
-Cw <= Vw: clamp Cx to [Vx, Vx + Vw - Cw]
-Cw >= Vw: clamp Cx to [Vx + Vw - Cw, Vx ]
-```
-
-垂直方向的法则完全相同。这支持了真正的焦点摆放,比如“源上的 `(0.8, 0.35)` 这个点应当出现在 Frame 的中心”,同时不会允许意外露出未被覆盖的边条。
-
-常见的九种对齐只是 `framePoint === contentPoint` 时的简写。居中 contain 和居中 cover 两边都是 `(0.5, 0.5)`。
-`top-left` 两边都是 `(0,0)`。如果作者刻意要让内容跑到外面,或者刻意让 Placement Frame 不被完全覆盖,就选择
-`free`;Runtime 绝不替他做这个创作决定。
-
-`FittedContent` 只包含推导出的 Content Frame,不重复上游血缘:
-
-```ts
-type FittedContent = {
- readonly contract: "svml.fitted-content@1";
- readonly contentFrame: SpatialFrame;
-};
-```
-
-当需要裁剪到 Placement Frame 时,对应的源采样窗口是由 Placement Frame 与 Content Frame 的交集机械推导出来的。它不是第三个由作者撰写的盒子。fit 之后的旋转或透视变换可能让采样边界不再是矩形,因此要放在这层基础布局之后应用。
-
-## 7. Fit 不是运动,也不是占用
-
-一次 fit 计算产出的是基础 Content Frame。缩放、平移、旋转、透视和入场运动会在之后对该结果做变换。它们不会改动 `ContentFit`,也不会改动源宽高比。
-
-这避免了旧有的歧义:`zoom` 被塞进 `contain`/`cover` 里,从而可能悄悄让所声明的 fit 不再成立。如果一段 Ken
-Burns 运动把一张 contain 的图片缩放到发生裁切,基础 fit 仍然是 `contain`,裁切是由之后的变换显式造成的。
-
-Spatial 的尺寸策略与固有时间占用也毫无关系。一段视频可以在空间上是 `contain`,同时在时间上播放一次、循环或拉伸。一张图片并不会仅仅因为它有固有的空间 extent 就获得某种固有的时间播放模式。
-
-## 8. 裁剪与 frame 形状
-
-Spatial 几何从不决定 Placement Frame 之外的像素是否可见。消费该几何的组件拥有显式的裁剪策略:
-
-```text
-none owned content may paint outside the Placement Frame
-placement-frame rectangular or rounded Frame clip
-owned-shape explicit shape/path owned by the same component
-```
-
-圆角半径、椭圆/路径蒙版、描边和阴影属于表现,不属于坐标。Text 的辉光完全可以合理地溢出它的 Placement
-Frame;Media 卡片通常会把每个局部样本都裁剪到同一个圆角 Placement Frame。Runtime 不会根据内容类型推断出任何默认值。
-
-自有形状只能裁剪同一个组件贡献内的元素。它不能蒙版或采样另一个同级 Track。
-
-## 9. 透明、纯色、模糊与替代底衬
-
-Placement Frame 中未被 Content Frame 绘制的区域本身不具备空间语义。Media/Text 的表现层提供的是一组有序的、显式自有图层。
-
-### 透明
-
-不产出任何底衬图层。透明像素会在 Composition 时自然露出更低的 Track。这不是跨 Track 采样:该 Track 在那里根本没有贡献像素。
-
-### 纯色或渐变
-
-一个 frame Paint 图层填满 Placement Frame。它没有源输入、没有 IntrinsicExtent、也没有 Content Frame。
-
-### 同源的模糊副本
-
-同一个源被显式地第二次连接进来,成为另一个局部样本:
-
-```text
-same source Artifact ─┬─> cover ContentFit -> blur/tone -> backing layer
- └─> contain ContentFit -> foreground layer
-```
-
-两个样本共享同一个 Placement Frame,但拥有各自独立的 Content Frame 和效果。Artifact 收集会对完全相同的字节去重;渲染仍然执行两次显式采样。模糊实现必须在裁剪之前做边缘扩展/超采,以免高斯核造出非预期的透明或发暗边框。
-
-### 另一张图片或另一段视频
-
-替代源通过自己的图边进入,并获得属于自己的 Inspection、IntrinsicExtent、ContentFit,对于有时长的媒体还包括占用策略。它不是藏在某个 Style Recipe 里的字符串路径。
-
-因此通用的 Media 作者模型是有序的局部图层,而不是特殊的 `foreground/background` 字段:
-
-```svml
-<media:Item id="portrait-card" frame={card} clip={rounded-card}>
- <media:Paint recipe={studio.media.card-base}/>
- <media:Layer source={portrait} fit={cover} effects={blurred}/>
- <media:Layer source={portrait} fit={contain}/>
-</media:Item>
-```
-
-把第二个源换成另一张图片,表达的就是替代底衬。删掉前两个子节点,表达的就是透明底衬。当需要与同级 Track
-交错时,每个图层都可以下降为自己的绝对堆叠 Present;作者层的 Track 绝不会变成隐式的堆叠上下文。
-
-完整的 Item/Sequence 作者模型见 [`media-track.md`](./media-track.md)。独立的深度堆叠集合模型见
-[`deck-track.md`](./deck-track.md)。两者都复用显式图层与边的法则,而不会把 Deck 变成 Media 的一种模式。
-
-### 仍然被禁止的做法
-
-对已经累积出来的 Composition 施加 `backdrop-filter` 不属于上述任何一种情形。那会让结果依赖于任意的下层
-Track。要模糊另一个视觉物,组件必须显式接收那个视觉/媒体。要模糊一个已完成的下层合成,必须由更高阶的组件先显式拥有并物化那个合成。
-
-## 10. Text、Ranking 与 Caption 的用法
-
-Spatial 提供公共几何,同时不强加单一布局模型:
-
-- Point Text 消费一个 `SpatialPoint`,并从精确的文字 extent 推导其 Content Frame;
-- Area Text 消费一个 Placement Frame,然后在其内部执行包自有的段落排流;
-- Path Text 消费一条显式的 `SpatialPath`;
-- Media 消费一个 Placement Frame,并为每个样本推导一个 Content Frame;
-- Ranking 消费一个 Placement Frame,并在其内部拥有 board、slot 和暂存区布局;
-- Comment Sticker 消费一个 Placement Frame,并拥有自己的头像/文字/内部布局;
-- Caption 可以消费一个 Placement Frame,但 Cue/Atom 的折行仍归 Caption 所有。
-
-Text 的布局盒和 Media 的 Content Frame 并不会仅仅因为都是矩形就成为同一个语义 Type。它们可以使用同样的纯几何辅助函数,同时各自仍然是包自有的已解析事实。
-
-## 11. 图数据流与不传播元数据
-
-```text
-CanvasSpace + Frame Projection
- │
- ▼
- SpatialFrame ───────────────────────────────┐
- │
-Media Inspection -> display extent -> IntrinsicExtent│
- ├─> fit -> FittedContent
-authored ContentFit ─────────────────────────────────┘ │
- ▼
- package Paint/motion/clip -> VisualTrack
-```
-
-每一条依赖都是图上的一条边。`BlobArtifact` 不会长出摆放字段。`SpatialFrame` 不会长出源、角色、时间或
-Artifact 摘要。`VisualTrack` 不会长出 `focal`、B-roll 或 Text 字段。亲和性由多输入的 fitting/lowering
-Operation 校验,而不是靠把血缘元数据复制到每一个值上。
-
-## 12. 保留与废弃的历史结论
-
-保留:
-
-- 归一化的作者摆放,以及感知源宽高比的锚定摆放;
-- 相互独立的源 fitting 与时间占用;
-- contain、cover、fit-width 和 fit-height 的方程;
-- 相互独立的前景采样与底衬采样;
-- 源焦点对齐与显式吸附点;
-- 可选的透明、纯色和自身模糊底衬;
-- 叠加在采样/摆放之上的局部运动。
-
-废弃:
-
-- 把区域、VLM 推断、避让、路径信号和固定盒子混在一起的联合体 `SpaceLocator`;
-- 把九宫格代号当作彼此独立的算法;
-- 用同一份 `focalX/focalY` 同时充当源点和目标点;
-- `MediaBoxStyle` 把“恰好一个前景加一个特殊底衬”当成通用模型;
-- 把 zoom 嵌进 fit;
-- 把渲染器专属的 CSS 与 FFmpeg 公式当成两份真相来源;
-- 隐式的 `object-fit`、`object-position` 或 `backdrop-filter` 默认值。
-
-## 13. 实现与验收顺序
-
-1. **已实现:** 新增聚焦的 `@narratage/spatial` 包,含 `CanvasSpace`、Frame/Point/Path、
- `IntrinsicExtent`、`ContentFit` 以及纯粹的 fit 校验;
-2. **已实现:** 把 Canvas 声明从 Film 装配中拆出,并把同一个 Canvas 值连接到 Track 布局与 Composition;
-3. **已实现:** 让一个简单的 Text Area 和一张简单的 Media 静态图片走通共享几何;
-4. **已实现:** 为每一条 fit 与对齐法则补上精确的浏览器测试;
-5. **已实现:** 验证有序的 透明/纯色/自身模糊/替代源 Media 图层;
-6. **已实现:** 让四个 Ranking 组件全部走显式的共享 Frame;Comment Sticker 在拥有自己的设计之前继续搁置;
-7. 只有到这时才冻结 Spatial 的 Type 以及相关的 Visual IR 行为。
-
-测试矩阵必须包含:
-
-- 竖版、横版和方形的固有 extent,分别对上竖版、横版和方形的 Frame;
-- 每一种尺寸模式,包括显式形变与 scale-down;
-- 全部九种等点对齐,外加源/frame 焦点不相等的情形;
-- bounded 与 free 对齐的对比;
-- 为负的、部分越界的以及完全在画布之外的 Placement Frame 与 Content Frame;
-- 父 Frame 百分比与像素偏移;
-- 在进入 Spatial 之前已归一化的源旋转/sample aspect;
-- 透明、纯色、同源模糊,以及替代图片/视频图层;
-- 相互独立的底衬 Content Frame 与前景 Content Frame;
-- 由所选渲染器实现锁定的取整/光栅行为;
-- 共享 Type 中不存在 Core、Runtime、Film 族、Text 族或 Media 族的判别字段。
-
-纯几何矩阵与真实 Chromium witness 现已覆盖每一种尺寸模式、全部九种等点对齐、不相等的 bounded/free 焦点,以及部分/完全越出 Canvas 的 Frame。Media 归一化还会在推导 `IntrinsicExtent` 之前,把容器旋转和非方形 sample
-aspect 物化成方形像素的视觉。Ranking 现在消费同样的显式 Frame 几何,且没有增加族判别字段。最终的 Visual IR
-兼容性审查现已通过;Spatial 自己的作者 Surface 仍处于预发布状态,在此期间不应再生长出任何包内局部的百分比词汇。
diff --git a/docs/zh/guide/spec/track-authoring.md b/docs/zh/guide/spec/track-authoring.md
deleted file mode 100644
index e59bf5eb..00000000
--- a/docs/zh/guide/spec/track-authoring.md
+++ /dev/null
@@ -1,386 +0,0 @@
----
-title: SVML 视频 Track 作者模型
-description: 官方 Track 包共享的可执行时间权威。聚焦的。
----
-
-# SVML 视频 Track 作者模型
-
-状态:官方 Track 包共享的可执行时间权威。聚焦的 `@narratage/temporal` 包实现了严格的 occurrence 展开、精确的有理数窗口投影,以及触发式的同级排程。共享的 Spatial、Text、Ranking、Media、Audio 和 Screen Overlay
-现已分别在 [`spatial-layout.md`](./spatial-layout.md)、
-[`typography-track.md`](./typography-track.md)、[`ranking-track.md`](./ranking-track.md)、
-[`media-track.md`](./media-track.md)、[`deck-track.md`](./deck-track.md)、
-[`audio-track.md`](./audio-track.md) 和
-[`screen-overlay.md`](./screen-overlay.md) 中给出规范。独立设计的
-`@narratage/comment-sticker` 包现在使用同一套 Temporal 与 Spatial 基础,并下降为一个同级 VisualTrack。这里刻意不构成对已发布公开 ABI 的兼容承诺。
-
-## 目的
-
-官方视频包需要一套共享的纪律,但不需要一个万能的 Track Program。Text、Media、Ranking 和 Comment Sticker
-拥有各不相同的作者语义,但它们不得各自发明互不兼容的方式来回答同样的时间问题。
-
-本文管辖的是包级 Program 及其确定性下降。它不会给 Core、Runtime、Composition 或某个 Provider 增加视频语义。公开的终端窄腰仍然是 [`VisualTrack` 和 `AudioTrack`](./track-composition.md)。
-
-完整的包作者关注点分离是:
-
-```text
-content
-temporal source
-occurrence expansion
-window projection
-optional sibling sequencing
-intrinsic-material occupancy
-spatial source and layout
-presentation
-```
-
-这些是相互独立的作者关注点,不是每个包都必须填写的字段,也不是公开 Track 合同上的字段。
-
-## 1. 时间流水线
-
-```text
-Narrative Selection ──> { start, end } occurrences ─┐
-Narrative Moment ─────> { cue } occurrences ────────┼─> occurrence expansion
-Program singleton ────> { program.start/end } ──────┘
- │
- ▼
- Temporal Window Projection
- │
- ▼
- raw directed candidate windows
- │
- ▼
- ProgramSpace intersection + validation
- │
- ▼
- frame-exact projected windows
- │
- ▼
- optional package-owned sibling sequencing
- │
- ▼
- final allowed windows
- │
- ▼
- intrinsic-material occupancy / playback
- │
- ▼
- render intervals + source-time sampling
-```
-
-各层回答的是不同的问题:
-
-| 层 | 问题 |
-|---|---|
-| anchor location | 某个作者写下的锚点位于哪一个物理帧? |
-| occurrence expansion | 一个作者项恰好消费一个 occurrence,还是消费每一个 occurrence? |
-| window projection | 这个 occurrence 可以占用哪一段 ProgramSpace 区间? |
-| sibling sequencing | 相关的项之间如何交接、重叠或彼此关闭? |
-| occupancy | 一个带固有时长的产物如何在自己的最终窗口内对自身采样? |
-
-任何 `full`、`fixed`、`relative`、`from`、`until`、`fill` 或 `fit` 联合体都不得再次把这些层压平。作者
-Surface 可以提供可读的简写,但它必须下降到这个分离后的模型。
-
-## 2. Temporal source 提供的是点,不是可渲染的窗口
-
-一个 Selection occurrence 贡献一对有序的已定位点:
-
-```ts
-type LocatedSelectionOccurrence = {
- readonly id: string;
- readonly start: FramePoint;
- readonly end: FramePoint;
-};
-```
-
-一个 Moment occurrence 贡献一个已定位点:
-
-```ts
-type LocatedMomentOccurrence = {
- readonly id: string;
- readonly cue: FramePoint;
-};
-```
-
-Program source 是一个单例 occurrence。每个投影环境还拥有 `program.start` 和 `program.end` 这两个精确的
-ProgramSpace 边界。
-
-Selection 的定位必须保持 Script 中的 occurrence 顺序,以及两个作者锚点各自的身份。它不得对这些点排序、合并、裁剪、互换或取绝对差值。一个 `end` 早于 `start` 的已定位 Selection 仍然是有效的点证据。只有消费者才会创建窗口,因此只有投影层才能判定它实际消费的那些点是否构成有效区间。
-
-组件自有的精确时间产物,比如已经定好时间的 Caption Cue,可以直接进入后面的帧窗口阶段。它不需要假装自己的时间来自某个 Narrative Selection。
-
-## 3. 点表达式与 Temporal Window Projection
-
-共享的投影代数在概念上是:
-
-```ts
-type TemporalPointExpression =
- | { readonly ref: "program.start"; readonly offset?: Duration }
- | { readonly ref: "program.end"; readonly offset?: Duration }
- | { readonly ref: "selection.start"; readonly offset?: Duration }
- | { readonly ref: "selection.end"; readonly offset?: Duration }
- | { readonly ref: "moment.cue"; readonly offset?: Duration }
- | { readonly ref: "absolute"; readonly at: Duration };
-
-type TemporalWindowProjection = {
- readonly start: TemporalPointExpression;
- readonly end: TemporalPointExpression;
-};
-```
-
-序列化合同将为时长使用显式单位。秒、毫秒和整数帧是三种不同的作者输入;换算到有理数 ProgramSpace 必须有唯一的确定性边界量化规则。`base_start` 和 `base_end` 是被废弃的名字,因为当前架构没有享有特权的 Base Track。
-
-常见的作者语义都只是普通表达式:
-
-| 语义 | 投影 |
-|---|---|
-| 整个 Program | `[program.start, program.end]` |
-| 恒等 Selection | `[selection.start, selection.end]` |
-| 从 Selection 开始起三秒 | `[selection.start, selection.start + 3s]` |
-| Selection 结束之后三秒 | `[selection.end, selection.end + 3s]` |
-| 从 Selection 开始一直持续 | `[selection.start, program.end]` |
-| 提前两秒引入到 Selection 开始 | `[selection.start - 2s, selection.start]` |
-| 某个 Moment 之后三秒 | `[moment.cue, moment.cue + 3s]` |
-| 最开始的三秒 | `[program.start, program.start + 3s]` |
-| 绝对时间的高级窗口 | `[absolute(2.5s), absolute(6s)]` |
-
-`full`、`manual`、`fixed` 和 `relative` 可以继续作为创作 UI 里的表现用词,但它们不是彼此不同的编译期时间法则。
-
-### 语义绑定可以刻意忽略局部坐标
-
-作者可以把一个项绑定到覆盖开场几个词的 Selection,同时把它投影到 `[program.start, program.start + 3s]`。即使该投影没有用到 Selection 已定位的 start/end 坐标,这个 Selection 仍然贡献了作者语义、occurrence 身份、基数和依赖真相。
-
-这在 `one` 之下是合法的。如下文所定义,它在 `each` 之下被刻意限制,因为重复一个与 occurrence 无关的投影只会制造出一模一样的窗口。
-
-## 4. Occurrence 展开是严格的,且先于投影
-
-旧的形态 `{ kind: "single", cardinality: "exactly_one" }` 把同一个事实说了两遍。可执行的合同只有两个互不重叠的选项:
-
-```ts
-type OccurrenceExpansion =
- | { readonly kind: "one" }
- | { readonly kind: "each" };
-```
-
-`each` 指的是编译期展开,不是媒体循环。作者 Surface 通过各自显式的 occurrence 设置暴露它,并下降为这个精确的值。
-
-### `one`
-
-`one` 遵循如下精确法则:
-
-1. 定位完整的源 occurrence 集合;
-2. 在求值投影之前,要求其基数恰好等于一;
-3. 拒绝零个或多于一个的 occurrence;
-4. 绝不选择 `first`、`last`、`best`、最近或最早者;
-5. 把那唯一的 occurrence 恰好投影一次。
-
-忽略 Selection/Moment 坐标的投影通常与 `one` 搭配。仅依赖 Program 的时间使用固有的单例 Program source,因此同样下降为 `one`,且不需要作者可见的基数设置。
-
-### `each`
-
-`each` 遵循如下精确法则:
-
-1. 要求至少有一个已定位的源 occurrence;
-2. 保持源 occurrence 的身份与源顺序;
-3. 为每一个源 occurrence 恰好实例化一个投影后的 occurrence;
-4. 在每个 occurrence 各自的局部点环境中独立求值同一个投影;
-5. 由作者项 id 与 occurrence id 推导稳定的结果身份,绝不使用偶然的排序下标;
-6. 绝不合并、去重、按时间排序、挑选优胜者或悄悄丢弃某个 occurrence;
-7. 只要有任何一个 occurrence 无法产出有效窗口,就让整个确定性 Operation 失败。
-
-当源 occurrence 多于一个时,投影的两个边界中至少有一个必须引用局部点(`selection.start`、`selection.end`
-或 `moment.cue`)。只使用 Program 点或绝对点的投影与 occurrence 无关,在 `each` 之下会被拒绝:它会造出若干语义不同、物理窗口却完全相同、彼此毫无时间区分的实例。作者应当改用 `one`、绑定 Program 单例,或分别写出显式的多个项。
-
-投影是一对一的。一个 occurrence 不能扇出成多个窗口。想从一个 Selection 或 Moment 得到两个窗口的作者,要么写两个项,要么使用一个显式构造这两个项的包自有组件。
-
-集合聚合不是第三种通用展开模式。可见性掩码或另一个确实以集合为值的组件,可以定义包自有的集合消费者,但普通的可视项不得借这个例外找回隐式的合并、并集或去重行为。
-
-Caption 的 `Mute` 是一个具体的包自有例子。它在 `CaptionProgram` 内部对显式引用的
-`CaptionDisplayWordSubset` 取并集;它不改变通用的 occurrence 展开,不制造投影窗口,也不进入 Core。
-
-### 基数不决定重叠
-
-`each` 只说明一个 occurrence 集合如何展开。它不说明产生的窗口是否可以重叠、相邻窗口是否应当合并,也不说明某个项是否应当在下一个项的 cue 处关闭。这些问题在投影之前无法回答,而且还可能取决于包的语义。
-
-投影之后,消费方的包必须为自己的组件选择并校验一种显式关系,例如:
-
-- `independent`:重叠的窗口是有意义的,彼此保持独立;
-- `disjoint`:同一个作者项的各 occurrence 之间出现任何重叠都是错误;
-- 由包自有的 Sequence 或 Deck 策略推导最终的同级窗口与交接。
-
-这种关系不属于 `OccurrenceExpansion`,也不由 Core 拥有。官方的简单 Text、Comment Sticker 以及类似的重复项,通常应当要求投影窗口互不相交。Media 组、Ranking 组件可以拥有更强的显式同级规则。任何包都不会获得隐藏的全局自动拼接行为。
-
-### 触发式阶段属于同级排序,不是另一种 temporal source
-
-Ranking、Tier、幻灯片以及其他渐进式组件,往往只有一组有序的触发 Moment。到达触发点 `i` 就进入状态 `i`;到达触发点 `i + 1` 就交接给下一个状态。这种情况使用的是同一套点定位与窗口投影。它不构成在 Core 里引入 `Stage`
-Type 或再加一种 Script 标记的理由。
-
-仅有 `N` 个触发点无法界定 `N` 个有限阶段:最后一个阶段仍然需要一个终止边界。因此触发式组件拥有一个显式的外层窗口 `[B, O)` 和一个显式的阶段终止点 `T`。`T` 可以引用 `O`,但包不得默认它一定如此。把两者分开,可以让 Ranking
-组件在 `T` 处结束最后一项,并让完全落定的 board 一直可见到 `O`。给定已定位并已量化到帧的触发点
-`p1 ... pN`,包可以推导出:
-
-```text
-stage 1 [p1, p2)
-stage 2 [p2, p3)
-...
-stage N [pN, T)
-settled suffix [T, O) (when T < O)
-```
-
-`p1` 之前的区间是非活动的,而不是隐式的第零阶段。如果某个设计需要一个从外层起点到 `p1` 的初始可见状态,那个状态必须显式写出,并会产生 `N + 1` 个阶段。渲染器绝不能因为存在 board 或背景就推断出它。
-
-这个模式有两种相关但不同的包自有下降方式:
-
-```text
-outer component [B, O)
-cumulative item 1 [p1, O)
-cumulative item 2 [p2, O)
-cumulative item 3 [p3, O)
-
-exclusive stage 1 [p1, p2)
-exclusive stage 2 [p2, p3)
-exclusive stage 3 [p3, T)
-settled suffix [T, O) when T < O
-```
-
-累积形态就是对 `[moment.cue, outer.end]` 做普通的 `each` 投影,外加一条显式的重叠策略。它天然实现了 Tier
-board:`p2` 之后,第 1 项和第 2 项同时处于活动状态;可见的前缀本身就是当前的逻辑状态。不需要物化任何 board
-快照。
-
-互斥形态从同样的候选出发,再施加一条后继交接策略:每个候选的最终结束成为下一个候选的开始,最后一个在 `T` 处结束。这可以驱动一个 Ranking 暂存区,让它恰好展示一个当前项,同时已落定的项作为独立的累积前缀保留。当
-`T < O` 时,后缀没有当前项,展示的是完成状态。包可以把这两个视图下降为同一个同级 `VisualTrack` 中相互独立的
-Present。
-
-触发式排序遵循如下精确法则:
-
-1. 保持作者/源的 occurrence 顺序与稳定的 occurrence 身份;
-2. 绝不按物理时间对触发点排序;
-3. 要求至少一个触发点、一个显式外层窗口和一个显式阶段终止点;
-4. 在帧量化之后,要求 `B <= p1 < p2 < ... < pN < T <= O`;
-5. 原子性地拒绝逆序触发点、同帧碰撞、落在外层窗口之外的触发点,以及位于 `T` 处或其之后的最后一个触发点;
-6. 只有当组件合同明确这样规定并要求基数相等时,才把有序的项列表与触发 occurrence 逐一对齐;
-7. 由作者项身份与 occurrence 身份推导项/阶段身份,绝不使用数组下标;
-8. 把累积重叠或后继交接保持为已声明的包语义,绝不交给 Runtime 选择。
-
-包也可以为每个子项各配一条单 occurrence 的 Moment 边。两种作者写法都下降为同样的有序同级候选;按位置对齐不是隐式的全局行为。除非另有组件确实需要消费它们,触发式阶段只是内部的确定性下降事实。Composition 收到的仍然只是普通的 Present。
-
-## 5. 投影校验与负区间
-
-对每一个展开后的 occurrence,投影按如下顺序执行:
-
-1. 在精确的有理数 ProgramSpace 中求值原始 start 和原始 end;
-2. 拒绝缺失、非有限或类型非法的点表达式;
-3. 若原始 end 早于原始 start,在裁剪之前就拒绝这个逆序窗口;
-4. 若原始 end 等于原始 start,拒绝这个零长度渲染窗口;
-5. 把方向正确的原始窗口与 `[program.start, program.end)` 求交;
-6. 确定性地把它量化为半开的整数帧跨度;
-7. 拒绝空结果或短于一帧的结果;
-8. 只返回完全落在 ProgramSpace 之内的 `{ startFrame, endFrameExclusive }` 值。
-
-裁剪到 ProgramSpace 是与唯一可观测渲染域求数学交集,而不是猜测时间。因此中间出现负值的点是允许的:
-
-```text
-cue = 1s
-[cue - 3s, cue + 2s] = [-2s, 3s]
-intersection [0s, 3s]
-```
-
-原始的逆序区间绝不能被裁剪、排序或取绝对值掩盖过去。完全落在 ProgramSpace 之外的区间会变成空并失败;它不会被悄悄略过。渲染器只收到已经校验过的帧跨度,不做任何时间修补。
-
-Selection 的一对交叉锚点本身并不构成投影失败。恒等投影 `[selection.start, selection.end]` 在产出逆序窗口时失败,而 `[selection.start, program.end]` 可以依然有效,因为它并不消费那个交叉的结束点。
-
-## 6. 固有时长与占用
-
-每一个被呈现的产物,要么拥有已验证的固有时长,要么没有。未知时长不得冒充其中任何一种;媒体检查必须先确定它,基于固有时长的占用策略才能执行。
-
-无时长产物 —— 例如 Text、静态图片、Comment Sticker 和普通的 Ranking UI —— 占满自己最终投影出的整个窗口。局部的入场、出场和循环动画属于该窗口内部的表现,不是合成出来的素材时长。
-
-带固有时长的产物会收到一个最终允许窗口和一条显式播放策略。概念上的可视策略是:
-
-```ts
-type IntrinsicPlayback =
- | { readonly mode: "once"; readonly align: "start" | "end" }
- | { readonly mode: "hold"; readonly align: "start" | "end" }
- | { readonly mode: "loop"; readonly align: "start" | "end" }
- | { readonly mode: "stretch" };
-```
-
-`once` 取代了含义含糊的旧词 `finish`。end 对齐绝不表示倒放。它表示正向播放的源在投影窗口的结束处结束。如果源比窗口长,start 对齐取源的头部,end 对齐取源的尾部。
-
-| 策略 | 源短于窗口 | 源长于窗口 |
-|---|---|---|
-| `once/start` | 从窗口开始播放,随后消失 | 播放源的头部直到窗口结束 |
-| `once/end` | 延迟出现,并在窗口结束处结束 | 播放源的尾部,并在窗口结束处结束 |
-| `hold/start` | 播放完后保持最后一帧 | 播放源的头部直到窗口结束 |
-| `hold/end` | 先保持第一帧,再播放到窗口结束 | 播放源的尾部直到窗口结束 |
-| `loop/start` | 从源头部相位开始重复 | 循环到放不下时裁掉 |
-| `loop/end` | 选择使源尾部落在窗口结束处的循环相位 | 同样的尾部相位法则 |
-| `stretch` | 把源的全部时间放慢塞进窗口 | 把源的全部时间加快塞进窗口 |
-
-Audio 必须在同一套排程之上定义自己如实的词汇。它不能像视频帧那样把一个 PCM 采样保持住。一次性音频之外是静音,补齐必须显式,循环就是循环,拉伸必须遵守作者声明的速度上下限。静态图片和无时长 UI 不接受无意义的固有播放模式。
-
-源的 trim 与源入/出点在被加入后,作用于占用之前,用来定义有效的固有源区间。Spatial 的 contain/cover 和 Text
-的 flow/shrink 作用于时间占用之后,与之无关。
-
-## 7. 包的归属
-
-原先那个大一统的 `contracts` 包被有意取消。名义上的共享语义由 Narrative、ProgramSpace、Media 和 Composition
-等聚焦模块各自拥有;Core 没有任何视频领域联合体的注册表。
-
-本文中的时间代数由聚焦的视频领域包 `@narratage/temporal` 实现。它拥有投影类型、validator 以及纯粹的帧窗口投影。它可以消费 Narrative 的点引用、SemanticMap 定位和 ProgramSpace,但它不得知道 Text、Media Track、
-Ranking、Comment Sticker、Film、HyperFrames、Runtime 或某个 Provider。
-
-官方包随后都使用同一份实现:
-
-```text
-Text ───────────────┐
-Media Track ────────┤
-Deck Track ─────────┤
-Ranking ────────────┼─> @narratage/temporal ─> frame windows
-Comment Sticker ────┤
-Screen Overlay ─────┤
-Audio Track ─────────┘
-```
-
-Caption 那份组件自有的定时 Cue 投影可以直接进入最终的帧窗口阶段。Speech Spine 确立 ProgramSpace 和连续的语音
-take。其他 Track 可以通过两个结构锚点直接寻址整个 Narrative Segment,而不必假装它是一个作者写下的 Selection。
-
-## 8. 迁移影响
-
-当前的 `typography-track` Surface 使用共享时间包来处理 Program、Selection 和 Moment 绑定、显式点表达式,以及
-`one`/`each` 展开。它不再维护第二套时间枚举。Media Track 用同一个包处理 Item 与 Sequence 的时间。
-
-可执行的验收覆盖包括:
-
-- Selection 与 Moment 的 `one` 成功用例及基数失败用例;
-- Selection 与 Moment 的 `each`、稳定的 occurrence 身份以及源顺序的保持;
-- 拒绝在多 occurrence 情况下使用与 occurrence 无关的 `each` 投影;
-- 所有常见的点表达式组合,包括 Program 边界和绝对时间;
-- 交叉的源锚点在不同投影下分别被消费和被忽略;
-- 负偏移、ProgramSpace 求交、完全越界的窗口、逆序窗口、零长度窗口以及一帧的最小长度;
-- 有理数帧率与确定性的半开帧量化;
-- 多个 `each` occurrence 中有一个非法时的原子失败;
-- 包自有的 disjoint、independent 和 sequence 关系与展开保持相互独立;
-- 累积式与后继交接式的触发序列,包括显式终止边界、同帧碰撞、非单调的源顺序、稳定身份以及可选的显式初始状态;
-- 无时长产物,以及在每一种受支持的播放与对齐策略下更短/相等/更长的固有时长产物;
-- 显式的音频行为,而不是复用视觉的 `hold` 语义。
-
-## 9. Ranking 的特化
-
-完整的 Ranking 审查、组件拆分、组排程和可执行迁移见 [`ranking-track.md`](./ranking-track.md)。Ranking 现已验证:通用的时间代数足以支撑累积状态、互斥的当前阶段以及最终的落定后缀,而无需向 Core 或公开 Track 合同引入
-Ranking 或 `Stage` 语义。
-
-## 10. 迁移状态
-
-Comment Sticker 的内容、布局与表现已作为独立的作者包实现;它只共享通用的 Temporal、Spatial 和终端 VisualTrack
-词汇。目前已规范的 Track 家族中,没有任何一个仍被这套通用作者模型卡住。
-
-对目前已规范的每一个官方 Track 包,已实现的验收矩阵证明 Text、Caption、Media、Deck、Ranking 和 Screen Overlay
-都下降为同级 VisualTrack 值,而通用 Audio 以及可选的 Media/Ranking 声音都下降为同级 AudioTrack 值,且不改动
-Core 或 Composition。
-
-通用 Audio Track 和自包含的 Screen Overlay 现已分别在 [`audio-track.md`](./audio-track.md) 和
-[`screen-overlay.md`](./screen-overlay.md) 中单独规范。Audio 复用同一套时间代数,但拥有如实的采样占用语义,而不是视觉上的 hold 语义。Screen Overlay 是一份满画布的同级贡献,不是跨 Track 的调整图层。
-
-统一的 Item/Sequence Media 创作、全部运动通道、内部交接、显式音频投影以及 Speech Spine 受限的复用,现已在
-[`media-track.md`](./media-track.md) 中规范。旧的深度堆叠 Deck 是一个独立的高阶 Track 组件,见
-[`deck-track.md`](./deck-track.md)。B-roll 是一种剪辑用的 Recipe/用法,而不是另一个公开的 Track 家族。
diff --git a/docs/zh/guide/spec/track-composition.md b/docs/zh/guide/spec/track-composition.md
deleted file mode 100644
index 34343ee1..00000000
--- a/docs/zh/guide/spec/track-composition.md
+++ /dev/null
@@ -1,92 +0,0 @@
----
-title: SVML Track 与 Composition
-description: 可执行的仓库内部 `@1` 兼容窄腰;所有门禁均已通过。
----
-
-# SVML Track 与 Composition
-
-状态:可执行的仓库内部 `@1` 兼容窄腰;[`track-expressiveness.md`](./track-expressiveness.md) 中的所有门禁均已通过,公开发布仍然是另一件事。
-
-## 法则
-
-除了共享的 ProgramSpace 与画布几何之外,进入 Composition 的每一份视听贡献都是同级 Track。Composition 负责校验、排序和合并 Track。它不认识 Caption、Speech、Media、Text、Seedance 或任何作者包家族。
-
-```text
-Speech visual ───────────────> VisualTrack ─┐
-Media ───────────────────────> VisualTrack ─┤
-TimedCaption + style/cues ───> VisualTrack ─┤
-Text + style ────────────────> VisualTrack ─┼─> Composition
-Vignette / overlay ──────────> VisualTrack ─┤
-Speech audio / BGM / SFX ────> AudioTrack ──┘
-```
-
-## ProgramSpace
-
-ProgramSpace 拥有时长和精确的有理数帧率。它的时长必须结束在整数帧边界上。Track 不会把 ProgramSpace 的摘要复制进每一个值。在两者关系需要被校验的地方,图会显式连接 Track 与 ProgramSpace;validator 会拒绝落在所提供空间之外的帧窗口。
-
-画布宽度、高度和 clear color 是 Composition 的结构性事实。任何视频 Track 都不定义时长,也不会成为享有特权的
-Base。
-
-## VisualTrack
-
-一个 VisualTrack 是由某个作者包拥有的、自包含的渲染贡献。它显式绑定
-[`svml.visual-ir@1`](./visual-ir.md) 终端语言,并拥有零个或多个帧精确的 `VisualPresent` 值;每个 Present
-拥有自己的绝对 `(order, tieBreak)` 堆叠键,以及一棵自包含、不含代码的元素树,其素材来自 box、text、普通媒体和带类型的可合成 Surface 原语。父引用是 Present 局部的。媒体和精确字体通过内容寻址的 Artifact 引用进入,而不是通过 CSS URL 或环境字体名。每一个终端文本元素都带有非空的精确 Font 栈。它与 ProgramSpace 的身份关系是内在的,因为没有时钟就无法解释这个 Track。是哪些 Record 产生了它,属于
-Graph/Derivation 真相,被刻意排除在 Track 之外。
-
-Composition 会先把所有 Track 的 Present 拍平,再对它们排序。因此一个 Media 包可以在 z=30 拥有一块 board、在
-z=80 拥有一个图标,而来自另一个 Track 的 Text Present 位于 z=50。z 随时间变化由两个互不重叠、堆叠键不同的
-Present 表示。Track 的归属关系绝不会创建渲染堆叠上下文或来源链。
-
-每个元素都可以可选地携带帧精确的局部关键帧,作用于 opacity、transform、filter 或 clip-path。这些关键帧只作用于该 Present 的元素树。Media 包可以把一次成对转场下降为它自己两个 Present 上互补的动画,而不必让 Composition
-知道这个转场的名字或语义。
-
-VisualTrack 刻意不具备:
-
-- 家族或来源种类字段;
-- 对同级 Track 或 Present 的引用;
-- 任意脚本或全局 CSS 选择器;
-- 采样下层 Track 的 backdrop 滤镜或混合模式;
-- 跨 Track 的蒙版、求交或转场输入。
-
-当一个局部滤镜、裁切、变换、动画或转场只作用于该作者包显式拥有的素材,并下降为自包含的 Present 时,它就是合法的。需要两份原始素材的效果,必须在最终下降为 Track 之前显式收到这两份素材。任何 Present 都不得采样已经累积出来的下层合成。
-
-## AudioTrack
-
-一个 AudioTrack 拥有精确的 48 kHz 采样域片段、带类型的规范 WAV Artifact、源区间、循环相位、保持音高的播放速率、目标区间、增益,以及采样精确的淡入淡出。它没有视觉堆叠键,没有无实际作用的总线标签,也与 Speech Track
-没有任何特殊关系。语音音频、音乐、源声和音效使用同一份合同。
-
-## Composition
-
-Composition 包含一个画布和一组 VisualTrack/AudioTrack 值。ProgramSpace 是通往 Film 和最终渲染 Operation 的一条同级输入边,而不是被反复复制的血缘元数据。Composition:
-
-1. 依据由显式图边提供的 ProgramSpace 校验每一个 Track;
-2. 拒绝重复的 Track 身份和重复的视觉堆叠键;
-3. 拍平每一个 VisualPresent,并按绝对 `(order, tieBreak)` 挂载;
-4. 在 ProgramSpace 中混合 AudioTrack 片段;
-5. 为一个被显式选定的最终渲染包产出与渲染器无关的输入。
-
-Composition 不得按作者领域的家族做分支,也不得授予某个 Track 访问已累积下层像素的特权。因此全局 shatter、调整图层、Base FX 通道或 B-roll 垫底转场都不属于这个版本。也不会为 Base FX 保留任何专用占位。
-
-TimedCaptionProjection 仍然是一个中间语义值。Caption 的分组、样式和 cue 下降必须在字幕作为普通 VisualTrack
-进入 Composition 之前完成。
-
-## 最终渲染边界
-
-`@narratage/hyperframes` 是当前面向 Composition 及其唯一带版本 SVML Visual IR 的参考编译器。它绝不按 Speech、
-Caption、Media 或作者包身份做分支。它确定性地产出一份内容寻址的 `HyperframesDocument`,其 HTML 按绝对堆叠键交错排布各个 VisualPresent。它不得把一个创作层的 Track 挂载成一个孤立的视觉包装层。AudioTrack 的编译与最终
-mux 是同一个 ProgramSpace 之上相互独立的媒体操作;HyperFrames 是一个无声的可视目标。
-
-这不是一条排他的边界。未来的 `render-remotion` 或基于 API 的渲染包,可以消费同样的 Composition 与
-ProgramSpace,编译出另一份不可变的渲染器文档,并返回同一份媒体合同。Author Source 通过导入相应的包来选择这条路径;Core 和 Runtime 不会猜测渲染器。当前的示例显式选择 `render-hyperframes`。
-
-该文档必须(MUST)绑定精确的有理数帧率、为正整数的帧数、画布尺寸、Artifact 集合以及生成的 HTML。与
-ProgramSpace 的关系仍然显现在该 Producer 的输入边与 Derivation 上,而不是被复制进文档内容。它完整的视觉帧域是半开区间 `[0, frameCount)`。渲染第 `n` 帧必须只依赖这份不可变文档、它精确的 Artifact、被锁定的渲染器实现以及 `n`;它不得要求顺序求值第 `0..n-1` 帧。因此带状态的作者效果必须在进入这条边界之前物化成带类型的 Surface。
-
-帧分区属于 Runtime 策略。合规的 Endpoint 可以把该域切分成任意一组互不重叠的半开跨度,只要它们按序并起来等于
-`[0, frameCount)`,并在本地或远程求值这些跨度、对单个跨度重试,最后一次性拼装可视分块并 mux 音频。这些跨度不是作者声明、不是 Core Operation,也不是 Track 身份。已履约的渲染 Product 必须重复完全相同的 文档/帧域/画布绑定;亲和性会拒绝为另一个域拼装出来的结果。
-
-编译出的文档中包含的是 `svml-artifact://` 占位符,不是本地路径、API URL 或签名 URL。解析这些占位符是渲染前一刻的 Runtime/Provider 动作。因此同一份编译好的文档既能本地渲染也能远程渲染,而不改变作者意图或它的编译身份。精确文字会变成生成的 `@font-face` 规则,并禁用字体合成。带类型的 Surface 在同一条 Artifact 边界上保持所声明的尺寸、色彩空间、alpha 模式以及静帧/逐帧时间;它们不会从文件扩展名推断。渲染器会在暂存之前校验这些字节事实,否则拒收该文档。它的通用 Need Receipt 绑定被锁定的 Endpoint 实现,而 receipt 覆盖的元数据中那份渲染器专属证据,绑定实际用于布局的精确浏览器或不可变的远程部署。
-
-`@narratage/film` 把包级的任意元数装配实现为一次有限不可变的 TrackSet 折叠,随后是普通的 Composition。渲染是一个单独的、显式导入的下游 Fragment。Core 只收到定端口、单结果的 Operation,而以某个中间 Track 或 Composition
-为 Target 并不要求渲染器。Film Surface 必须下降到已实现的那个 Fragment,既不向 Core 增加 Track 家族,也不让某个 Track 依赖累积出来的同级像素。
diff --git a/docs/zh/guide/spec/track-expressiveness.md b/docs/zh/guide/spec/track-expressiveness.md
deleted file mode 100644
index d8f8bb55..00000000
--- a/docs/zh/guide/spec/track-expressiveness.md
+++ /dev/null
@@ -1,156 +0,0 @@
----
-title: SVML Track 表现力门
-description: E1–E8 全部门均可执行;仓库内部的 `svml.visual-track@1` 窄腰在公开发布前冻结。
----
-
-# SVML Track 表现力门
-
-状态:E1–E8 全部门均可执行;仓库内部的 `svml.visual-track@1` 窄腰在公开发布前冻结。
-
-## 目的
-
-一份公开的 Track 合同必须解决最终的视听合成问题,同时不假装所有作者包都共用同一套创作模型。Text、
-Caption、Media、Ranking 以及未来的包各自拥有自己的 Program,并且可以各自独立地修订这些 Program。它们只有在把包内专有含义 lower 成自包含、帧精确的贡献之后,才能进入 Composition。
-
-这道门只回答一个问题:
-
-> 生产中实际使用的文字、媒体盒、字幕、层叠与局部特效行为,能否在不向 Track、Composition 或 Core
-> 添加包族专有字段的前提下,lower 进同一条 Track 窄腰?
-
-通过这道门并不会产生一个统一的 `TrackProgram`。Track Program 的六条轴——内容、时间源、窗口投影、空间源、占位与呈现——仍然是给包作者的分离纪律,而不是六个强制的公开字段。共享的创作法则和当前的时间设计在 [`track-authoring.md`](./track-authoring.md) 中定义。
-
-## 三个绝不能塌陷的层级
-
-```text
-Package Program Resolved package facts Public Track
-─────────────────────── ──────────────────────────── ─────────────────────
-TextLayout / CaptionMode boxes, glyph runs, windows Presents
-Media sampling policy -> source-time and crop mapping -> exact frame spans
-effect recipe local animations absolute stacking
-semantic/spatial locator final placement self-contained elements
-```
-
-Package Program 是可编辑的作者真相。Resolved package facts 是编译器证据。公开 Track 是终端的视听贡献。
-Composition 绝不能从第三层反推出前两层。
-
-“扁平 Track”意味着不存在 Track 嵌套,也不存在隐式的 Track 层叠上下文。一个 Present 仍然可以拥有自己内部的 element 树。一条创作 Track 可以在互不相关的绝对层叠位置上输出多个 Present,从而让对等 Track
-能够穿插在它们之间。
-
-## 必需的见证
-
-### E1 —— Text 三盒 lowering
-
-官方面向浏览器的参考 lowering 必须能够独立表达:
-
-1. 一个外层 placement frame;
-2. 一个控制折行、对齐、方向和溢出的内层 layout box;
-3. 面向完整 frame、收缩包裹的内容、每一行或每一个词的 paint 目标。
-
-辉光或阴影这类 paint 溢出,不能仅仅因为它可见于内容盒之外,就变成行测量的输入。包可以通过把 paint
-挂到自己拥有的另一个 element 上来 lower 一个 paint 目标;公开 Track 绝不能因此获得
-`textBackgroundTarget` 字段。
-
-### E2 —— Caption range/cue/content lowering
-
-Caption 包必须能够 lower 出一个硬性的区间边界、一个 cue 级的对齐/动画容器、收缩包裹的行,以及可以独立设置样式或动画的词。双字体和逐词 paint 都是包内事实。Composition 只看到普通的 element 与 Present。
-
-### E3 —— Media 双帧 lowering
-
-media 包必须能够 lower 出一个已解析的 Placement Frame,其中包含同一个 Artifact 的两次独立采样。每次采样按 [`spatial-layout.md`](./spatial-layout.md) 推导自己的 Content Frame:
-
-- 一个背景采样,带自己的 cover/zoom/blur/tone 处理;
-- 一个前景采样,带自己的 contain/cover/fit-height 与焦点对齐。
-
-公开 Track 绝不能因此获得 B-roll、焦点预设或模糊背景字段。Artifact 依赖收集必须对共享的内容摘要去重。
-
-### E4 —— 独立的绝对层叠
-
-一个包可以在某个绝对层叠位置输出一块 board,同时在其他位置输出图标或标签。来自另一条 Track 的 Present
-必须能够出现在它们之间。Track 身份只是把一份自包含的渲染贡献归组;来源信息存在于 Graph/Derivation 中,而 Track 永远不构成渲染层叠上下文。
-
-### E5 —— 局部 motion 与交接
-
-帧精确的 opacity、transform、filter 与 clip 动画,可以作用在某一个 Present 所拥有的 element 上。同时拥有交接双方素材的包,可以 lower 出互补的 Present。任何普通 Track 都不得采样其下方已累积的像素,也不得修改同级 Track。
-
-Media 包还必须额外证明:进入、维持、退出、逐层采样 motion 与 Sequence 配对交接,始终是分离的通道并具有固定的合成顺序。独立的 Depth-Stack Deck 包必须证明其集合状态的重排能与 Card 局部 motion 组合,且不会把
-Deck 含义带进终端 Track。完整模型在 [`media-track.md`](./media-track.md) 和
-[`deck-track.md`](./deck-track.md) 中定义。
-
-### E6 —— 物化的视觉兜底
-
-如果某个组件的确定性视觉结果无法用参考 element 树表达,它必须能够把自己拥有的结果物化,并作为带类型的
-Surface 贡献出来。为了实现透明穿插,最终的公开合同必须绑定足够的 surface 信息,以区分可合成的带 alpha
-视觉与不透明视频。仅仅命名一个 `.webm` Artifact 并不能证明这一性质。
-
-### E7 —— 通用音频 lowering
-
-官方 Audio 包必须能把任意显式的归一化源、源裁切、一次性播放、循环、有界的保持音高拉伸、起止对齐、增益与淡入淡出,全部 lower 进同一条对等的 `AudioTrack` 窄腰。重叠的 clip 混合时不做优先级裁剪,也没有特权的 Base 声道。容器流选择、响度处理与 Provider 放置仍然是上游或下游图中显式的工作。完整边界在
-[`audio-track.md`](./audio-track.md) 中定义。
-
-### E8 —— 自包含的屏幕叠加
-
-全画布特效必须 lower 成普通的绝对层叠 Present,由自己拥有的 Visual IR element 或自己拥有的带 alpha
-Surface 构成。它不能读取已累积的下层合成结果,不能预留最高 z-index,也不能引入合成后阶段。因此无源的模糊、调色和真实的 zoom-blur 都会 fail closed;它们诚实的形态是在 Screen Overlay 包之外消费显式的媒体或
-Surface 输入。参见 [`screen-overlay.md`](./screen-overlay.md)。
-
-## 旧系统攻击矩阵
-
-| 历史行为 | 包所拥有的真相 | 必需的终端见证 |
-|---|---|---|
-| Text 的 placement/layout/paint 盒 | Text Program | 自己拥有的嵌套 element |
-| frame/content/line/word 背景 | Text Program | paint 挂在所选的自有层级上 |
-| caption 的 range/cue/content 盒 | Caption Program | 自己拥有的嵌套 element 与词级局部动画 |
-| 双字体、中日韩、emoji 与书写方向 | Text/Caption Program | 文本 element 加锁定的字体/布局依赖 |
-| 媒体内容盒加前景/背景采样 | media 包 | 引用同一个 Artifact 的同级媒体 element |
-| 焦点裁切与 fit-height | media 包 | 确定性的采样/放置结果 |
-| 全画幅 B-roll、角落 GIF 与下方媒体卡 | Media Item + Recipe | 同一个 Item 对不同 Frame 的 lowering |
-| 显式的媒体替换组 | Media Sequence | 协同的自有 outgoing/incoming Present |
-| 可见的卡片历史与深度重排 | Depth-Stack Deck Track | 建立在普通 Present 之上的包内集合状态 |
-| 处在互不相关 z 值上的 Ranking board 与图标 | Ranking Program | 多个绝对层叠 Present |
-| Media 的进入/退出与配对转场 | Media Program | 覆盖自有 Present 的局部帧精确 keyframe |
-| 任意的自包含视觉 | component/provider | 物化的可合成 Surface Artifact |
-| 音乐/音效/额外人声的放置 | Audio Program | 普通的帧精确 AudioTrack clip |
-| 闪白/暗角/颗粒/蒙层 | Screen Overlay Program | 自有的全画布 Present 或 alpha Surface |
-| 对一张画面做模糊/调色 | 显式输入的媒体特效 | 变换后的自有素材,绝不是背景采样 |
-
-## 当前的可执行证据
-
-`packages/hyperframes/test/track-expressiveness.test.ts` 构造的是包内私有的 lowering 见证,而不是往公开合同里加 Text、Caption、Media 或 Ranking 字段。已有的包测试另外证明了普通的 Text、Caption 与 Media
-Program 都经由同一条 Track 与 Composition 路径 lower。
-
-当前的 `VisualTrack / VisualPresent / VisualElement` 候选显式绑定 `svml.visual-ir@1`,能够编码结构性的三盒、两盒、词级局部和绝对层叠等情形。有两条通用的终端事实现在是可执行的,而不是包族补丁:
-
-1. `FontArtifactRef` 把每个精确的 fallback 字面、weight 和 style 绑定到一个或多个 Blob 源。
- Unicode-range 分片仍然算一个逻辑字面;精确的文本 lowering 会为每个源输出按内容寻址的
- `@font-face` 规则,并禁用字体合成;
-2. `CompositableSurfaceRef` 把尺寸、sRGB 色彩空间、不透明/直通 alpha 语义以及静帧/逐帧时间绑定到 Blob
- 字节。带动画的 Surface 必须与所在 Present 的帧数和 ProgramSpace 速率完全一致。
-
-`packages/hyperframes/test/browser-visual.test.ts` 还会驱动真实的 Hyperframes 浏览器帧:已安装的按内容寻址开放字体通过有序的 fallback 栈绘制拉丁、中日韩与 emoji/符号字形;Text 证明了 Point/Area/Path 布局、全部 Box 目标、富文本 run、水平/垂直排版流、clip/ellipsis/有界收缩、有序 Paint、局部 Mask 以及正/反向的选择器时钟;Fine Caption 证明了多行折行、完整字形 Paint、每一种字形卡拉 OK 模式以及 joined 拖尾 Pill 的行几何;DepthStack Deck 证明了显式的新旧集合状态重排、精确标签以及单 worker 与分区渲染逐字节一致的帧;全部十一个官方 Screen Overlay 组件在顺序渲染与并行帧渲染下保持完全相同的像素;一张 50% 直通 alpha 的
-PNG 在蓝色 Track 上合成,得到预期的像素值。这个 opt-in 的宿主测试通过 `SVML_BROWSER_TESTS=1` 运行;它证明的是当前本地渲染器路径,而不是未来所有的托管 Runtime。
-
-IR 的 style 词汇是封闭的,而不是任意 CSS:组件无法在不改变协议版本的前提下加入未知属性、依赖环境的值或另一种浏览器语言。现在每个终端文本 element 都携带非空的精确 Font Artifact 栈;系统字体兜底和原型字族都是非法的。Runtime 在通用 Receipt 中绑定其锁定的实现。HyperFrames 还会把精确的本地浏览器摘要,或所配置的远程渲染器部署摘要,记录在受 receipt 覆盖的元数据中。共享的、基于 ffprobe 的 Surface 字节校验,会在本地暂存之前证明所声明的尺寸、时间、SDR/sRGB 与 alpha 事实;不具备这一能力的 Endpoint 会拒绝携带
-Surface 的文档。
-
-E1–E5、E7 和 E8 现在都有各自的官方包见证。`@narratage/typography-track` 在不引入公开 Text 字段的前提下证明了完整的三盒/Text 终端模型;Caption 和 Media 证明了它们各自独立的包模型。`@narratage/audio-track`
-证明了精确的 48 kHz 源/目标映射、重叠法则、有界拉伸以及一份共享的本地/远程 `AudioProgramPlan`;
-`@narratage/screen-overlay` 证明了只使用自包含的自有像素、显式种子、扁平层叠,以及对下层合成类特效的
-fail-closed 拒绝;`@narratage/deck-track` 证明了在新旧可见 Card 并集之上的纯绝对帧集合重排,同时复用了
-Card 局部的 Media lowering。`@narratage/ranking` 提供了真正的 E4 见证:board、stage 与独立层叠的 Item
-与一条无关的对等 Track 相互穿插,同时四个 Ranking 组件在分区浏览器渲染下都保持帧纯净的渐进状态。E6
-则通过带类型的路径、独立安装的非原生 Text 见证、字节校验器以及 fail-closed 的 Endpoint 能力检查得以完成。
-
-## 冻结标准
-
-仓库内部的 Track v1 窄腰满足以下全部条件:
-
-1. E1–E8 都具备可执行的结构、HTML lowering 与音频 plan 见证;
-2. 官方的 Text/Caption lowerer 要么绑定精确的字体 Artifact,要么显式物化其字形结果,并且用于布局的
- 渲染器实现是受 receipt 绑定的;
-3. E6 具备一条带类型的可合成 Surface 路径,其 Runtime 会校验所声明的媒体事实并记录渲染器实现;
-4. 安装一个新的 Track 包,不需要改动 Core、Film、Composition 或族注册表;
-5. 公开的 Track 数据中不含任何 Caption、Text、Media、Ranking 或 provider 判别字段;
-6. 一个包可以修订自己的 Program schema,同时只需保持或版本化自己的 lowerer;
-7. 不受支持的跨 Track 采样 fail closed。
-
-因此该实现对后续的仓库包工作已经冻结。公开的 npm 发布、非受信代码隔离与最终的发布版本号仍然是独立的发布门;它们不会重新打开这条窄腰的视频含义。
diff --git a/docs/zh/guide/spec/typography-track.md b/docs/zh/guide/spec/typography-track.md
deleted file mode 100644
index 278efb24..00000000
--- a/docs/zh/guide/spec/typography-track.md
+++ /dev/null
@@ -1,372 +0,0 @@
----
-title: Typography Track 创作与表达力
-description: 官方二维 Text 包的已实现预发布权威文档。
----
-
-# Typography Track 创作与表达力
-
-状态:官方二维 Text 包的已实现预发布权威文档。完整声明的模型与浏览器证据都通过冻结的仓库内部
-Track/Visual IR 窄腰执行;作者 Surface 本身尚不是已发布的 ABI。
-
-## 1. Conclusion
-
-完整的 Text 模型不是一个更大的 `TextAppearance` 对象。它是下列各维度的正交乘积:
-
-```text
-authored document
-× temporal binding
-× spatial form
-× text flow
-× exact typography
-× ordered Paint layers
-× deterministic local motion
-```
-
-一个普通的透明标题、整句话外面套一个背景、每一行渲染行各套一个背景、孤立的词 Pill、连成一体的荧光笔几何,以及逐字素簇的打字机效果,都是这个模型中的不同取值。它们不是不同的渲染器组件,也不需要在 Core、
-Film 或 Composition 里开分支。
-
-“完整”不等于把有史以来所有特效都塞进一个扁平对象。它的意思是:
-
-1. 每一种常见的二维编辑器 Text 行为,在这个模型里都有带类型的落点;
-2. 新增 Style 和 Recipe 不会改动共享协议;
-3. 输入或物理规律确实不同的能力,使用显式的兄弟组件;
-4. 任何超出终端元素语言表达范围的自包含视觉效果,都可以物化为带类型的 compositable Surface,而无需改动
- Core。
-
-官方 Text 包拥有这套作者语义。SVS 只提供具名的不可变 Recipe 参数。Visual IR 仍是共享的终端视频语言。
-HyperFrames 只是这门语言的一个渲染器,而不是 Text 的定义。
-
-## 2. Ownership and package boundary
-
-常规二维能力仍然只用一个包:
-
-```text
-@narratage/typography-track
- Point Text
- Area Text
- Path Text
- Text Style / Inline Style declarations
- Text Program validation
- temporal/spatial resolution adapters
- VisualTrack lowering
-```
-
-Point、Area 和 Path 之所以是三个独立的作者组件,是因为它们断言的几何不同。它们不是那种会让端口忽隐忽现的 `mode` 字段。它们仍然可以下降到同一个带标签的 `TextItemProgram` 联合类型,并作为对等 item 共存于同一个
-Text Track 中。
-
-有两项能力的输入实质不同,因此留在常规 Text 之外:
-
-- Text Mask 组件必须显式消费文字形状和它所遮罩的媒体/图形;它不得去采样恰好位于其下方的那个 Track;
-- 真正的 3D Text 拥有挤出、材质、灯光和相机语义,通常物化为带 alpha 的 Surface。
-
-在后续的官方发行版中,它们可以是独立的包,也可以是需要单独导入的组件。包的数量是发行决策;但它们的图合同必须保持分离。两种情况都不会给 Core 增加分支。
-
-官方的本地 Text Mask 见证刻意只接受一个精确的单行 Area Text 形状和一个显式的静态 `CompositableSurface`。富 run、多行/Path 排布、序列动画和定时素材都会 fail closed,并走与 3D Text 相同的“物化 Surface”逃生路线。这条边界避免了假装浏览器 `foreignObject` 遮罩是可移植且精确的。
-
-## 3. Authored Text Document
-
-Text item 消费的是一份有界的富文档,而不是 HTML 片段或任意树:
-
-```ts
-type TextDocument = {
- readonly paragraphs: readonly TextParagraph[];
-};
-
-type TextParagraph = {
- readonly id: string;
- readonly runs: readonly TextInline[];
- readonly style?: TextParagraphStyleRef;
-};
-
-type TextInline =
- | { readonly kind: "text"; readonly text: string; readonly style?: TextInlineStyleRef;
- readonly language?: string; readonly direction?: "auto" | "ltr" | "rtl" }
- | { readonly kind: "break" };
-```
-
-具体的公开名称在实现过程中可能改变,但下列法则不会变:
-
-1. 文字、空白、标点和显式换行完全保持作者顺序;
-2. 一个纯字符串是“一个段落、一个 run”的简写;
-3. Span 只改变可显式继承的 typography/Paint,绝不改变时序或 item 摆放;
-4. 只有当被选用的布局/Paint/运动操作确实需要单元时,源文字才会被切分;
-5. 切分具备 Unicode 与 locale 意识,并绑定到具体实现;
-6. 任何 Text 包都不得用 LLM 去改写、拆分或装饰作者内容;
-7. `uppercase`、小型大写字母及类似变换属于表现层,绝不改变被声明的文档值。
-
-确定性的单元层级是:
-
-```text
-document > paragraph > explicit run > rendered line > Unicode word > grapheme cluster
-```
-
-`rendered line` 只有在确定了精确字体和最终几何之后才存在。字素簇是常规情况下作者可见的最小动画/装饰单元。一个 shaped glyph 未必与某个 Unicode 字符或字素簇一一对应,因此公开的作者模型不得假装它们一一对应。
-
-## 4. Three spatial forms
-
-时间源与窗口投影仍由 [`track-authoring.md`](./track-authoring.md) 规定。Text 不会发明 `full`、`from`、
-`until` 或另一套时序系统。
-
-空间摆放与 Text 排布相互独立。[`spatial-layout.md`](./spatial-layout.md) 中的共享模型提供解析后的
-point/frame/path 几何;Text 随后用三种形态之一去解释这份几何。
-
-### 4.1 Point Text
-
-Point Text 是锚定在某个解析点上的固有尺寸文字。它通常紧贴内容、不做软换行,并使用显式的内联/块级锚点。手动分段仍然有效。常规的标题、标签和贴纸都是 Point Text。
-
-### 4.2 Area Text
-
-Area Text 在解析出的 frame 内排布。它显式选择:
-
-- 内联与块级尺寸:在有意义处使用 `hug | fixed`;
-- 四条逻辑边的内边距;
-- 内联对齐:`start | center | end | justify`;
-- 块级对齐:`start | center | end`;
-- 换行:`none | word | grapheme`;
-- 溢出:`visible | clip | ellipsis | shrink`;
-- 截断/缩放策略下可选的最大行数;
-- 书写方向与书写模式;
-- 可选的分栏数与栏间距;
-- 是否裁剪到摆放 frame;
-- 度量边:line box、cap height 或 ink bounds。
-
-溢出是诚实的作者意图。`ellipsis` 可能省略掉可见的被声明文字,因此必须显式写出。`shrink` 必须声明最小缩放比例,并在完整文字仍放不下时失败;它不得悄悄越过那个下限。这一点与 Caption 不同——Caption 绝不能裁剪或丢弃作者的显示 Atom。
-
-### 4.3 Path Text
-
-Path Text 消费一条显式自有的矢量路径,外加 Text Document 和 Style。它拥有路径侧、朝向、起止边距、对齐、反向和溢出。给边距做动画会让文字沿路径移动;给路径做动画则改变路径本身。它绝不会去发现或采样另一个
-Track 中的形状。
-
-Path Text 需要一个通用的矢量/路径终端原语,或者一个自有的物化 Surface。它不能靠序列化一段无类型的
-SVG/HTML 字符串来声称已实现。
-
-## 5. Typography
-
-Typography 与 Paint、布局相互分离,但参与布局度量。它包括:
-
-- 精确的有序 `FontStackRef`,环境字族名称仅作为显式的原型路径;
-- 字号、字重与 normal/italic/oblique 字形;
-- 可变字体轴;
-- OpenType 特性选择与字体合成策略;
-- 字偶距、字距、词间距与行高;
-- 语言、方向与书写模式;
-- 基线偏移、制表宽度、缩进与段间距;
-- 跟随字形度量的下划线/上划线/删除线几何;
-- 文字变换、小型大写字母、上标与下标的呈现;
-- 所选布局实现支持时的 CJK 专用间距/压缩控制。
-
-精确的字体字节仍是显式的作者图输入。它们不是藏在 Recipe 里的名字,也不由 Runtime 挑选。字体 shaping/布局实现的身份由被接受的推导/渲染 receipt 绑定;Core 仍然对字体一无所知。
-
-## 6. Ordered Paint model
-
-Paint 是一个有序列表,而不是一组互斥的 CSS 简写。Text Program 拥有这些带类型的 Layer:
-
-```ts
-type TextPaintLayer =
- | { readonly kind: "fill"; readonly paint: ColorPaint }
- | { readonly kind: "stroke"; readonly paint: ColorPaint; readonly widthPx: number;
- readonly placement: "inside" | "center" | "outside" }
- | { readonly kind: "shadow"; readonly paint: ColorPaint; readonly offset: Vector;
- readonly blurPx: number; readonly spreadPx: number }
- | { readonly kind: "glow"; readonly paint: ColorPaint; readonly blurPx: number;
- readonly spreadPx: number }
- | { readonly kind: "box"; readonly target: TextPaintTarget;
- readonly continuity: "isolated" | "joined"; readonly decoration: BoxDecoration };
-```
-
-`ColorPaint` 支持纯色、线性渐变和径向渐变,带一份有序的颜色与不透明度 stop 列表。重复使用 Stroke 或
-Shadow 是合法的,顺序具有语义。渲染器可以用复制字形 Layer 的方式来下降外描边,但那是实现细节,而不是一个含义随浏览器而变的作者开关。
-
-Box Paint 有这些目标:
-
-```text
-frame | content | paragraph | line | run | word | grapheme
-```
-
-它的装饰独立拥有填充、border、内边距、各角圆角和阴影。`isolated` 为每个目标创建一个独立的装饰单元。
-`joined` 在每一条渲染行内把相邻的被选片段合并,并给该行片段一个诚实的端头。非法组合(例如在 `frame` 上使用 continuity)会在 Text 校验期失败。
-
-这一条规则就表达了常见的全部情形:
-
-| 外观 | Paint 取值 |
-|---|---|
-| 普通透明文字 | 不加 Box Paint Layer |
-| 整句话一个胶囊 | `target=content` |
-| 整个摆放面板 | `target=frame` |
-| 每条渲染行一个框 | `target=line, continuity=isolated` |
-| 每个词一个 Pill | `target=word, continuity=isolated` |
-| 连成一体的马克笔/荧光笔 | `target=word, continuity=joined` |
-| 字母块 | `target=grapheme, continuity=isolated` |
-
-气泡尾巴是显式声明、附着在某个 Box Paint Layer 上的部件,带有方位、偏移、尺寸和自己的 Paint。它不会悄悄强制生成一个背景。毛玻璃不属于普通 Box Paint,因为它要采样 Track 背后的像素;它需要一个显式自有的媒体输入,或者一个物化的自包含组件。
-
-## 7. Motion and sequence selectors
-
-运动与 Style 相互独立,因此同一套外观可以有不同的入场、循环和退场。一个 item 可以拥有:
-
-1. item 级的有限关键帧,覆盖位置、缩放、旋转、斜切、不透明度、模糊和裁剪;
-2. Path Text 的路径边距动画;
-3. 一个或多个作用于文字单元的序列动画器。
-
-一个序列动画器是下列各项的乘积:
-
-```text
-unit selector × selected property channels × keyframes × stagger/order
-```
-
-单元选择器可以指向 paragraph、line、run、word 或 grapheme。它显式选择范围、正向/反向顺序、延迟/扩散、循环次数以及任何确定性随机种子。属性通道包括局部变换、不透明度、模糊和字形 Paint 值。打字机是这个模型上的一个字素簇显现预设,而不是字符串切片。滚动字幕和横向滚动只是 Area Text 上普通的 item 级运动。
-
-任何运动名称都不是隐藏的回调。`pop`、`spring`、`slide` 或 `wobble` 这类预设,会在终端渲染之前编译成有限的或在解析上有界的、由包自有的运动值。两条运动通道通过各自独立的包裹器或显式的变换合成顺序进行合成;后者不会覆盖前者。
-
-动画的生命期与 Item 的可见性相互独立。一段运动可以在 Item 窗口结束前就完成,此时它最终被声明的状态一直保持到 Item 消失。一段运动也可以超出窗口,此时 Item 窗口只是裁剪它,而不改写它的时序。序列运动和 Path
-Text 运动遵循同样的规则。这既避免了仅仅因为标题一直可见就拒绝一段完全合理的快速标题动画,也避免了为了塞进短窗口而悄悄给慢速动画重新计时。
-
-## 8. Style declaration, SVS and readable authoring
-
-声明与使用保持分离。SVS 仍是通用的标量 Recipe 语言;它不得学会 Text 的 Paint 数组,也不得变成第二个渲染器。Text Style Surface 读取具名 Recipe、精确的 Font 引用和重复的结构化 Layer 声明,校验它们,然后产出一份完整的、由包自有的 `TextStyle` Record。
-
-预期的作者写法是:
-
-```svml
-<text:Style id="plain" recipe={studio.text.plain} font={fonts.inter}/>
-
-<text:Style id="word-pills" recipe={studio.text.word-pills} font={fonts.inter}>
- <text:Box target="word" continuity="isolated" recipe={studio.text.word-pill}/>
-</text:Style>
-
-<text:Style id="poster" recipe={studio.text.poster} font={fonts.inter}>
- <text:Stroke recipe={studio.text.poster-outline}/>
- <text:Stroke recipe={studio.text.poster-keyline}/>
- <text:Shadow recipe={studio.text.poster-shadow}/>
-</text:Style>
-
-<text:Track id="titles" space={film.space}>
- <text:Point id="hook" style={poster} during={story.selection.hook} placement={hook-point}>
- Nothing hidden in the Runtime.
- </text:Point>
-
- <text:Area id="explanation" style={word-pills} during={story.selection.body} placement={body-frame}>
- <text:P>Every <text:Span style={plain}>visible word</text:Span> stays authored.</text:P>
- </text:Area>
-</text:Track>
-```
-
-item 也可以改为消费一个普通的图 `Text` 值:
-
-```svml
-<copy:Value id="headline">Nothing hidden in the Runtime.</copy:Value>
-
-<text:Track id="titles" space={film.space}>
- <text:Point id="hook" content={headline}
- style={poster} during="program" placement={hook-point}/>
-</text:Track>
-```
-
-`content={Text}` 与内联正文内容互斥。前者从精确的图值物化出一段纯文本文档 run;后者拥有有界的富文档,并可使用 `P`、`Span` 和 `Break`。这既阻止了 Frontend 把运行时 Text 复制进隐藏的被声明状态,又让富排版明确归本包所有。
-
-这套写法在当前共享的时间/空间 Surface 语法上可执行。它的重要性质是:
-
-- 作者导入并选择 Typography 包;
-- Style 先声明后使用,并且就是普通的被声明 Record;
-- 简单的 Style 值仍然紧凑;
-- 重复的 Layer 就是重复的声明,而不是编号字段或编码字符串;
-- 解释 Recipe 词汇的是 Markup Surface,不是 SVS;
-- 精确字体以及时序/摆放事实都走显式的图边;
-- 最终得到的 Text Program 里没有未解析的 Recipe,也没有 Runtime/provider 选择。
-
-## 9. Compiled pipeline and data gates
-
-```text
-Markup Surface
- (authored rich document | graph Text input) + named Style + temporal binding + spatial binding
- │
- ▼
-Text Program package-owned author truth
- │
- ├── located temporal points ──> projected frame windows
- └── resolved spatial geometry
- │
- ▼
-Resolved Text Layout
- exact fonts + shaping + wrapping + fragment geometry + motion schedule
- │
- ▼
-VisualTrack
- code-free terminal elements/keyframes or typed owned Surface
- │
- ▼
-Composition -> selected final renderer
-```
-
-每一项外部依赖都由图边承载。Text 值只包含固有的 Text 语义。任何上游 Record 都不会仅仅因为后面某个 Text
-消费者需要,就凭空长出 text、role、source、style、font 或 layout 元数据。
-
-布局实现可以使用被锁定的浏览器、HarfBuzz/Skia/Pango 或其他精确引擎。它必须绑定实际的实现和字体字节。浏览器的 line box 不会作为通用图元数据向外传播;它要么是渲染器 receipt 所拥有的确定性终端布局,要么是包自有的、用于物化 Surface 的已解析几何。
-
-## 10. Terminal Visual IR findings
-
-冻结的仓库内部 `svml.visual-ir@1` 只承载本次迁移所证明的最小增量:
-
-- `text-flow` 拥有有界富文档、精确 typography、有序 Paint 和序列值;
-- `path-text` 拥有带类型的矢量命令和精确的路径排布事实;
-- 本地 `mask` 原语恰好拥有一个终端遮罩源和一个内容根;
-- 有序可重复的 Paint 与 Unicode 单元动画始终是序列化数据,绝不是回调;
-- 精确的字体 Artifact 和带类型的 `CompositableSurface` 值走普通的 Artifact 边界。
-
-HyperFrames 参考编译器实现了这些原语,同时不向作者包暴露 HTML、CSS、SVG 或渲染器脚本。一个单独安装的非原生 Text fixture 证明了挤出/材质/灯光/相机语义可以留在窄腰之外,只贡献一个带类型的 alpha Surface。不受支持的富遮罩走同一条路径。
-
-这些始终是视频终端的事实,绝不是 Core 的事实。渲染器 receipt、Surface 字节校验、Deck/Ranking 见证以及最终的兼容性审计,如今都在不向共享窄腰添加 Text 语义的前提下通过。
-
-## 11. Feedback into Fine Caption
-
-Text 与 Caption 共享的是排版实现难题,而不是作者语义。
-
-下列内容可以通过一个聚焦的视频领域实现库共享(暂定为 `@narratage/typography`);在真正出现跨包边之前,它没有 Author Surface,也没有图 Type:
-
-- Unicode 字素簇/词/双向切分;
-- 精确的字体栈加载与 shaping;
-- 渐变 stop、有序的描边/阴影/辉光以及 box 装饰几何;
-- 行片段与 joined Pill 几何;
-- 字素簇安全的打字机效果与确定性的选择器求值;
-- 终端富文本下降辅助函数。
-
-这并不会造出一个通用的 Text Program。Caption 仍保留 `CaptionDisplaySequence`、Cue、Atom、语音对应关系和卡拉 OK 时序。Text 仍保留段落、富 run、point/area/path 布局和任意序列选择器。
-
-本次审计发现的、可落到 Caption 上的具体改进是:
-
-1. 允许有序且可重复的 Stroke/Shadow/Glow Layer,取代当前“每种只能一个”的上限;
-2. 允许任意渐变 stop,并支持径向渐变而不只是两个 stop 的线性渐变;
-3. 复用显式的内/中/外 Stroke 语义,而不是依赖渲染器的描边行为;
-4. 复用字素簇安全的文字变换/显现以及精确 shaping;
-5. 把 joined 的活动 Pill 几何保留在共享的行片段实现上。
-
-Caption **不得**继承 Text 的 ellipsis、裁剪、缩放适配、路径排布、任意富 run 时序或选择器生成的语音时序。它始终展示作者完整的不可变 Atom,并且只从已证明的 Atom 窗口激活。
-
-## 12. Migration order and acceptance gates
-
-实现是替换掉早期切片,而不是不断堆砌兼容字段:
-
-1. **已实现:** Text Document、Point/Area 带标签几何、Text Style 和有序 Paint 值;
-2. **已实现:** 精确字体、frame/content/paragraph/line/run/word/grapheme 的 Box Paint,以及富 run 静态下降;
-3. **已实现:** 共享的 Temporal Window Projection 和显式的 Point/Frame/Path 空间输入;
-4. **已实现:** 确定性溢出、方向/书写模式以及浏览器像素证据;
-5. **已实现:** item 运动与具备 Unicode 意识的序列选择器;
-6. **已实现:** Path Text,外加一个精确有界的本地 Text Mask 和 fail-closed 的物化边界;
-7. **已实现:** `CompositableSurface` 逃生路线,以及一个单独安装的非原生 Text 包见证;
-8. **已实现:** 在全仓库所有关卡通过之后冻结仓库内部的 Visual IR。
-
-验收至少要求下列各项的浏览器/渲染证据:
-
-- 透明 Point Text 与固定尺寸 Area Text;
-- content、frame、多行的 line、孤立 word、joined word 和 grapheme 各种框;
-- 跨换行边界的嵌套 run 样式;
-- 使用精确字体的拉丁文、CJK、RTL、组合符号和彩色 Emoji;
-- 可重复的外/内描边、多重阴影、线性/径向渐变以及辉光;
-- visible/clip/ellipsis/有界 shrink 四种溢出;
-- item 的入场/退场/循环,以及正向/反向的 grapheme/word/line 序列化;
-- Path Text;
-- 显式的 Text Mask 归属,以及一个物化的高级效果兜底;
-- 不修改 Core、Runtime、Film 或任何无关的 Track 包。
-
-已声明的 Text 包关卡现已全部通过。这确立了一个构建在冻结的仓库内部 Track/Visual IR 窄腰之上的完整预发布
-Text 包;npm 发布以及 Text 作者 Surface 自身的公开版本承诺仍是另外的事情。
diff --git a/docs/zh/guide/spec/visual-ir.md b/docs/zh/guide/spec/visual-ir.md
deleted file mode 100644
index 76989c50..00000000
--- a/docs/zh/guide/spec/visual-ir.md
+++ /dev/null
@@ -1,111 +0,0 @@
----
-title: SVML Visual IR
-description: 已实现的仓库内部 `@1` 兼容窄腰;尚未作为 npm ABI 发布。
----
-
-# SVML Visual IR
-
-状态:已实现的仓库内部 `@1` 兼容窄腰;尚未作为 npm ABI 发布。
-
-## 定位
-
-`svml.visual-ir@1` 是官方 SVML 视频栈共享的唯一一种无代码的终端视觉语言。它是:
-
-- 一份公开的视频领域协议;
-- 不是作者组件;
-- 不是 Provider API;
-- 不属于领域无关的 Core。
-
-Text、Caption、Media、Ranking 以及第三方视觉包可以各自持有互不相干的作者 Program。但在进入
-`Composition` 之前,每个包都必须把自己解析后的结果下降为一个显式指明本 IR 的 `VisualTrack`。渲染器
-adapter 校验并编译这门共同语言;它永远不需要知道来源是哪个组件家族。`@narratage/hyperframes` 是第一个参考 adapter,而不是协议的持有者。
-
-```text
-author Program shared terminal protocol final-render route
-───────────────────── ───────────────────────── ─────────────────────
-Text / Caption / Media ───> SVML Visual IR in Track ───┬─> HyperFrames
-custom visual producer ───> typed CompositableSurface ───┤
- ├─> future Remotion
- └─> future render API
-```
-
-把它称作组件“方言”是一种误导。当前视频栈只有一个官方目标协议。一个组件不可能引入另一套 CSS 语义,同时还声称自己产出的是同一份 IR。
-
-## 合同
-
-每一份 `svml.visual-track@1` 都携带:
-
-```text
-visualIr = svml.visual-ir@1
-```
-
-这个值先被携带该 Track 的 Record 覆盖,再经由 Composition 与所选渲染器文档的 Record 传递覆盖。Runtime
-不能用另一种视觉语言去重新解释一份已经编译完成的 Track。
-
-该 IR 只包含:
-
-- 带绝对堆叠键、帧级精确的 Present;
-- 每个 Present 局部的一棵由 `box`、`text`、`image`、`video` 或 `surface` 元素构成的树;
-- 内容寻址的媒体、字体与 Surface 引用;
-- 一份封闭的、面向浏览器的样式声明清单;
-- 按帧寻址、作用于 opacity、transform、filter 与 clip 的局部 keyframe;
-- 安全的 data/ARIA 属性。
-
-它不包含组件家族、作者 recipe、provider 名称、选择器、脚本、外部 URL、运行时凭据、同级 Track 引用,也不包含累积合成输入。
-
-## 封闭的渲染词汇
-
-实现导出 `VISUAL_STYLE_NAMES_V1`。不在该清单内的声明一律非法,即使某个特定的 Chromium 构建恰好能理解它。
-`position`、`display`、`overflow`、`object-fit`、direction 和 writing mode 这类关键的枚举型事实,还额外使用封闭值集。
-
-以下情况一律失败关闭:
-
-- 未知的 CSS 属性;
-- `position: fixed` 与滚动布局;
-- `backdrop-filter` 或 `mix-blend-mode` 的跨 Track 采样;
-- 用 CSS `url(...)` 代替带类型的 Artifact 引用;
-- `var(...)`、`env(...)` 或 `attr(...)` 这类依赖环境的值;
-- 会破坏声明结构的标点;
-- 对局部动画集合之外属性的动画。
-
-这有意不是“把任意 CSS 塞进 JSON”。这种 CSS 形状的表示,仅仅是被锁定浏览器目标的序列化指令词汇。
-
-## 校验归属
-
-校验有四个彼此独立的持有者:
-
-1. `@narratage/visual-ir` 持有封闭词汇与身份;
-2. `@narratage/composition` 持有 Track schema 与内在自包含性校验;
-3. Film 与最终渲染 Producer 通过显式图边接收 ProgramSpace,校验关系型的帧域事实,而不往 Track 值里添加
- 血缘字段;
-4. 被选中的本地或托管渲染 Endpoint 锁定实际的渲染器实现,并校验产出的媒体事实。
-
-Core 内不含任何视频专用分支。它的通用准入通路会调用已安装的 Type 持有者 validator,来校验内在的树、样式与 Artifact 事实。涉及另一个值的关系性校验,仍然是普通的多输入 Producer 检查。
-
-## 扩展法则
-
-安装一个新的视觉组件不得修改本协议。该组件只能二选一:
-
-1. 把自己的 Program 下降为已有的元素与样式;或者
-2. 把它所持有的视觉物化为一个带类型的 `CompositableSurface`,并贡献该 Surface。
-
-只有当某项能力既具备广泛复用价值,又无法通过上述两条路径表达时,才有理由改动 `svml.visual-ir@1` 合同。在 Narratage 处于预发布阶段期间,标识符保持为 `@1`,确切的实现身份由摘要和 package lock 承载。新增一种
-Text 样式、Caption 模式、Media recipe、转场或 Provider,本身永远不构成理由。
-
-因此,旧系统的 Text 三盒模型、Caption 的 range/cue/content 模型和 Media 的内容帧采样模型,都属于包自有的
-Program 与下降证据。它们是本 IR 表达力的见证,而不是被加进 IR 的字段。
-
-## 冻结证据
-
-标识符与封闭词汇都是可执行的。精确的 Text 字体、多行装饰、Path/Mask/overflow、Media 焦点采样和 alpha 现在都有真实的浏览器证据;一个单独安装的非原生 Text 包可以穿过带类型的 Surface 窄腰,而不改动 Core、Film 或
-HyperFrames。仓库内部的 `@1` 窄腰之所以被冻结,是因为以下四道彼此独立的闸门现在都能真正执行:
-
-1. 通用的 Need Receipt 绑定被锁定的 Endpoint 实现/配置/Runtime closure,同时 HyperFrames 那份被 receipt
- 覆盖的 attestation,在本地绑定确切的浏览器字节,在远端绑定显式配置的不可变渲染器部署摘要;
-2. 共享的媒体执行器在本地渲染之前解码 Surface 字节,并证明尺寸、帧时序、SDR/sRGB 兼容性以及
- opaque/straight-alpha 事实;不具备该 verifier 的渲染器 Endpoint 必须拒绝携带 Surface 的文档;
-3. Deck 与 Ranking 提供其余的包自有 Track 表达力见证;
-4. 全仓库的边界审查证明 Composition 中不含任何包家族/provider 判别字段,且终端编译器中不存在隐藏的跨 Track
- 采样通路。
-
-这为本仓库后续的包开发冻结了该协议。它并不意味着 npm 包已经发布、每个作者 Surface 都已冻结,也不意味着未来的公开发布可以跳过分发与安全审查。
diff --git a/packages/caption-fine/README.md b/packages/caption-fine/README.md
index 597e72e1..20688f7b 100644
--- a/packages/caption-fine/README.md
+++ b/packages/caption-fine/README.md
@@ -62,5 +62,4 @@ One Track handles the default and ordered Style replacements. `font=` accepts ei
the compact `fonts:Stack` above. Family, weight and style exist only on those exact font values;
the Recipe owns size and appearance. Fallbacks keep their own honest face metadata. Omitting the stack is invalid rather than an environment-font
prototype path. Fine never clips author text and intentionally has no `max-lines`.
-Common Caption, Composition and Core know none of Fine's Recipe fields or layout policy. See
-[Fine Caption Style](https://narratage.hypit.ai/guide/spec/caption-fine) for the full parameter surface, defaults and non-goals.
+Common Caption, Composition and Core know none of Fine's Recipe fields or layout policy.
diff --git a/packages/core/README.md b/packages/core/README.md
index e3f78e17..813fd2c3 100644
--- a/packages/core/README.md
+++ b/packages/core/README.md
@@ -25,8 +25,7 @@ receipt uniformly but never executes the validator; trusted Host admission lives
It generically verifies Surface declarations in the same immutable module closure, but never
parses or dispatches one. It does not parse source, resolve package locations, execute
implementations, access artifact bytes, select Provider endpoints or discover undeclared graph
-structure. Those responsibilities remain outside Core. The normative laws are in
-[Core Kernel](https://narratage.hypit.ai/guide/spec/core-kernel).
+structure. Those responsibilities remain outside Core.
For a non-video domain, the irreducible reusable base is only `@narratage/protocol` plus `@narratage/core`.
Most source languages will also use `@narratage/elaborator` to turn modular component declarations into a
diff --git a/packages/media-track/README.md b/packages/media-track/README.md
index bf651bec..b2b467dc 100644
--- a/packages/media-track/README.md
+++ b/packages/media-track/README.md
@@ -20,5 +20,4 @@ restricted internal projection.
The automatic lowering is Surface behavior, not a privileged Runtime route: it produces the same
ordinary Media Pipeline Operations that an author can write explicitly through
`@narratage/media-pipeline`. The package does not add Media, B-roll or Provider meaning to Core,
-Film, Composition or HyperFrames. “B-roll” is an editorial use of an ordinary Item or Sequence. The
-complete pre-release contract and acceptance matrix live in [Media Track](https://narratage.hypit.ai/guide/spec/media-track).
+Film, Composition or HyperFrames. “B-roll” is an editorial use of an ordinary Item or Sequence.
diff --git a/packages/ranking/README.md b/packages/ranking/README.md
index cb662d04..bbacfdfa 100644
--- a/packages/ranking/README.md
+++ b/packages/ranking/README.md
@@ -34,5 +34,3 @@ they define the board vocabulary rather than dynamic item copy.
<ranking:ColumnItem label={winner}/>
</ranking:Column>
```
-
-See [Ranking Track](https://narratage.hypit.ai/guide/spec/ranking-track) for the normative pre-release model.
diff --git a/packages/spatial/README.md b/packages/spatial/README.md
index 44a97c9c..6b18cc7b 100644
--- a/packages/spatial/README.md
+++ b/packages/spatial/README.md
@@ -5,5 +5,4 @@ and deterministic two-frame content fitting. It owns no timing, Paint, motion, m
renderer, Provider or Core behavior.
The package exposes self-described `Canvas`, `Point`, `Path`, `Extent`, `Frame`, `AnchoredFrame` and
-`AspectFrame` author Surfaces plus pure geometry functions and fixed-port Producers. See
-[Shared Spatial Layout](https://narratage.hypit.ai/guide/spec/spatial-layout).
+`AspectFrame` author Surfaces plus pure geometry functions and fixed-port Producers.
diff --git a/packages/temporal/README.md b/packages/temporal/README.md
index c33670c8..5b32366b 100644
--- a/packages/temporal/README.md
+++ b/packages/temporal/README.md
@@ -6,5 +6,3 @@ expressions, intersects them with ProgramSpace and returns validated half-open f
The package also provides pure sibling-window validation and triggered-stage scheduling. It defines
no Core branch, renderer behavior, media playback policy, Provider or authoring super-program.
-
-The normative design is [Track Authoring Model](https://narratage.hypit.ai/guide/spec/track-authoring).