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.

refactor: fold speaker authoring into Seedance kits

kashorin6a005280645b
diff --git a/.gitignore b/.gitignore
index d2ca30d7..a54dc106 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,5 +1,37 @@
node_modules/
dist/
+coverage/
+.nyc_output/
+.DS_Store
+*.log
+*.tmp
+*.temp
+
+# Local editor and agent state is never part of the distribution.
+.claude/
+.codex/
+.agents/
+.idea/
+.vscode/
+
+# Runtime credentials belong to the selected credential store, never Git.
+.env
+.env.*
+!.env.example
+*.key
+*.pem
+*.p12
+*.pfx
+*credentials*.json
+*service-account*.json
+
+# Local databases and test-run reports are deployment state.
+*.db
+*.db-*
+*.sqlite
+*.sqlite-*
+playwright-report/
+test-results/
docs/.vitepress/cache/
docs/.vitepress/dist/
.svml-cache/
diff --git a/docs/.vitepress/config.ts b/docs/.vitepress/config.ts
index 3d5ba51a..7298c620 100644
--- a/docs/.vitepress/config.ts
+++ b/docs/.vitepress/config.ts
@@ -107,6 +107,7 @@ export default defineConfig({
"open-source-distribution.md",
"build-archive-and-egress.md",
"caption-gemini-provider-contract.md",
+ "demo-assets.md",
"graph-first-value-boundary.md",
"hyperframes-aws-runtime.md",
"image-transform.md",
@@ -119,7 +120,6 @@ export default defineConfig({
"runtime-adapter-loading.md",
"source-and-run-compilation.md",
"speech-alignment.md",
- "twinit-capability-migration.md",
],
markdown: {
theme: {
diff --git a/docs/README.md b/docs/README.md
index 612629f4..1f84c67a 100644
--- a/docs/README.md
+++ b/docs/README.md
@@ -32,6 +32,7 @@ the working tree contains only documents that should inform current implementati
Artifact retention and optional Host materialization as four separate concerns.
- [`speech-alignment.md`](./speech-alignment.md) — the Script and the recording as two
independent observations, and what locating guarantees.
+- [`demo-assets.md`](./demo-assets.md) — distribution boundary and provenance gate for documentation media.
- [`source-and-run-compilation.md`](./source-and-run-compilation.md) — mandatory Source Header,
peer Author/Run graphs, deterministic composition and compilation data gates.
- package READMEs under [`../packages`](../packages) — executable APIs and ownership.
@@ -62,12 +63,11 @@ the working tree contains only documents that should inform current implementati
- [`../spec/track-composition.md`](../spec/track-composition.md)
- [`../spec/track-expressiveness.md`](../spec/track-expressiveness.md)
- [`../spec/visual-ir.md`](../spec/visual-ir.md)
-- [`twinit-capability-migration.md`](./twinit-capability-migration.md) — one explicit ledger for every
- active Twinit node: complete replacement, expressible-but-missing Template, backlog or deliberate omission.
-The finite six-member Seedance Kit migration recorded in the ledger is complete as data-only Text
-Templates. Low-level Seedance invocation, Estimate, media utilities and final Blob flow remain
-shared and must not be reopened as Kit-specific execution code.
+The finite seven-member Seedance Kit family is complete as data-only Text Templates. Low-level
+Seedance invocation, Estimate, media utilities and final Blob flow remain shared and must not be
+reopened as Kit-specific execution code. Superseded migration ledgers live in Git history rather
+than the current documentation tree.
The shared `svml.visual-track@1` / `svml.visual-ir@1` terminal waist is frozen inside the repository
before publication. Individual author Surfaces remain pre-release unless their own specification
@@ -94,6 +94,8 @@ says otherwise.
- `docs/implementation-status.md` states repository reality.
- `docs/roadmap.md` states planned work and non-goals.
- package and example READMEs explain one executable unit.
+- `docs/public/` contains documentation media cleared for redistribution; its provenance rules are
+ tracked in `demo-assets.md`.
Do not preserve a superseded design in the current tree merely as a discussion log. Keep durable
decisions in the current architecture/specification, and rely on Git history for rejected drafts.
diff --git a/docs/demo-assets.md b/docs/demo-assets.md
new file mode 100644
index 00000000..ca647dc3
--- /dev/null
+++ b/docs/demo-assets.md
@@ -0,0 +1,12 @@
+# Demo assets
+
+`docs/public/` is copied byte-for-byte into the documentation site. It is therefore distribution
+content, not a test fixture, Build archive or ignored development output.
+
+Only brand-neutral assets cleared for repository and website redistribution belong in this
+directory. Generated Build outputs and example inputs belong under the ignored example
+`assets/` / `output/` paths.
+
+Every image, video, voice and likeness must have recorded provenance and redistribution permission.
+Replacing an asset must update its source mapping under
+`docs/.vitepress/theme/components/` and repeat that review.
diff --git a/examples/talking-film-golden/README.md b/examples/talking-film-golden/README.md
index 3f91daa8..c941c126 100644
--- a/examples/talking-film-golden/README.md
+++ b/examples/talking-film-golden/README.md
@@ -21,7 +21,7 @@ The source explicitly chooses:
It does not contain API keys, endpoint URLs, queue names, database names or deployment topology.
Those facts belong to a Runtime Profile. Runtime may bind `seedance.mini.*` to KIE, Volcengine or
-Hypit, but it may not reinterpret Seedance as Kling or silently replace Gemini cue planning with a
+a hosted deployment, but it may not reinterpret Seedance as Kling or silently replace Gemini cue planning with a
different model family.
Prompt declarations, media declarations and `.svs` recipes appear before their uses. The Script
@@ -97,7 +97,7 @@ bindings:
hyperframes.render@1: hyperframes.local
```
-A Hypit profile changes endpoint bindings, not `main.svml`. Existing Values, black frames and other
+A hosted profile changes endpoint bindings, not `main.svml`. Existing Values, black frames and other
preview realizations remain explicitly selected Candidates in a BuildRequest; they do not alter
the author source or teach Runtime to guess a creative method.
diff --git a/examples/talking-film-graph-check/svml.packages.lock b/examples/talking-film-graph-check/svml.packages.lock
index 2914f101..7bf23d81 100644
--- a/examples/talking-film-graph-check/svml.packages.lock
+++ b/examples/talking-film-graph-check/svml.packages.lock
@@ -594,7 +594,7 @@
{
"name": "@narratage/fonts-open",
"version": "0.0.0-dev",
- "digest": "sha256:2636ba4fa9323b3d5715db76f819ed702ff55fbc8a1e7f44f0401389f2a7bdb6"
+ "digest": "sha256:14f53b6b08cfa1a24d0e62c80755e9e24891de9fec7285bfca449019fd0cf965"
},
{
"name": "@narratage/generation",
@@ -654,7 +654,7 @@
{
"name": "@narratage/render-hyperframes",
"version": "0.0.0-dev",
- "digest": "sha256:18d5f3a63c71135771af5e9b47f8bcd6cf41662adcb3f1d1628dce6138f75931"
+ "digest": "sha256:5f30fd4c9ce4de8984408d125498f58b06899011d6e07b43bea6ff3498490fe1"
},
{
"name": "@narratage/run-markup",
@@ -749,7 +749,7 @@
{
"name": "@narratage/whisperx",
"version": "0.0.0-dev",
- "digest": "sha256:7209d52d61e9c9237b59695a30f2f04a055d83053f06439ba9e9a31dd6cd189c"
+ "digest": "sha256:7ee7f559fdc0508b6a7447838eb0a121cbe60c03522c6969638e0e8d3c04068f"
}
],
"packages": [
@@ -814,7 +814,7 @@
"version": "0.0.0-dev"
},
"facetsDigest": "sha256:31b51352edcb4f34beccc9d9ac8cf7397125a840971a72c912df3d34d1f3feda",
- "closureDigest": "sha256:cf404e70036afe899420d8271a0fa29e640fbe727fe25020c0a852c33db25cae"
+ "closureDigest": "sha256:31d17d4d2f19841ed32f85ac7f9ffe6e77b45b2514cd4cd52a96a979cc848ba0"
},
{
"specifier": "@narratage/generation",
@@ -886,7 +886,7 @@
"version": "0.0.0-dev"
},
"facetsDigest": "sha256:4210736adc699b3b5f86bf8b0cd84cd7a3c946754c8f19ceae5e0390a3d1d9c5",
- "closureDigest": "sha256:24acdfad8a63da6d5b9f3205e0bd04cb060fafea67b9c0d20623f108514644f0"
+ "closureDigest": "sha256:396057461f2715ea0a2890418e237c35a51ccc379ead2cef85392cdfb54015dd"
},
{
"specifier": "@narratage/run-markup",
@@ -1030,8 +1030,8 @@
"version": "0.0.0-dev"
},
"facetsDigest": "sha256:73fceef213e8d6bcad92532c5d712842e62921743b81bc7bb4d9022a92067359",
- "closureDigest": "sha256:e24bf6745248f12fd2fe968f693e7629800a61a6eafda464dd8c3b08c93215f5"
+ "closureDigest": "sha256:6866f06f64152df2bf378c2788d187cfc1718a83e7471104342ef6818d5da5e7"
}
],
- "digest": "sha256:abf08ab6f205657a5ec996cf9ef272e742ba0705db2cd7d62ae933bd14a234bd"
+ "digest": "sha256:22c0e0bb1449e0e44a60132912cae3552647861eccc85aacc957788d10df9489"
}
diff --git a/examples/talking-film-live/svml.packages.lock b/examples/talking-film-live/svml.packages.lock
index 29c29512..6ecebe31 100644
--- a/examples/talking-film-live/svml.packages.lock
+++ b/examples/talking-film-live/svml.packages.lock
@@ -599,7 +599,7 @@
{
"name": "@narratage/fonts-open",
"version": "0.0.0-dev",
- "digest": "sha256:2636ba4fa9323b3d5715db76f819ed702ff55fbc8a1e7f44f0401389f2a7bdb6"
+ "digest": "sha256:14f53b6b08cfa1a24d0e62c80755e9e24891de9fec7285bfca449019fd0cf965"
},
{
"name": "@narratage/generation",
@@ -659,7 +659,7 @@
{
"name": "@narratage/render-hyperframes",
"version": "0.0.0-dev",
- "digest": "sha256:18d5f3a63c71135771af5e9b47f8bcd6cf41662adcb3f1d1628dce6138f75931"
+ "digest": "sha256:5f30fd4c9ce4de8984408d125498f58b06899011d6e07b43bea6ff3498490fe1"
},
{
"name": "@narratage/run-markup",
@@ -754,7 +754,7 @@
{
"name": "@narratage/whisperx",
"version": "0.0.0-dev",
- "digest": "sha256:7209d52d61e9c9237b59695a30f2f04a055d83053f06439ba9e9a31dd6cd189c"
+ "digest": "sha256:7ee7f559fdc0508b6a7447838eb0a121cbe60c03522c6969638e0e8d3c04068f"
},
{
"name": "cmu-pronouncing-dictionary",
@@ -833,7 +833,7 @@
"version": "0.0.0-dev"
},
"facetsDigest": "sha256:31b51352edcb4f34beccc9d9ac8cf7397125a840971a72c912df3d34d1f3feda",
- "closureDigest": "sha256:cf404e70036afe899420d8271a0fa29e640fbe727fe25020c0a852c33db25cae"
+ "closureDigest": "sha256:31d17d4d2f19841ed32f85ac7f9ffe6e77b45b2514cd4cd52a96a979cc848ba0"
},
{
"specifier": "@narratage/generation",
@@ -905,7 +905,7 @@
"version": "0.0.0-dev"
},
"facetsDigest": "sha256:4210736adc699b3b5f86bf8b0cd84cd7a3c946754c8f19ceae5e0390a3d1d9c5",
- "closureDigest": "sha256:24acdfad8a63da6d5b9f3205e0bd04cb060fafea67b9c0d20623f108514644f0"
+ "closureDigest": "sha256:396057461f2715ea0a2890418e237c35a51ccc379ead2cef85392cdfb54015dd"
},
{
"specifier": "@narratage/run-markup",
@@ -1049,8 +1049,8 @@
"version": "0.0.0-dev"
},
"facetsDigest": "sha256:73fceef213e8d6bcad92532c5d712842e62921743b81bc7bb4d9022a92067359",
- "closureDigest": "sha256:e24bf6745248f12fd2fe968f693e7629800a61a6eafda464dd8c3b08c93215f5"
+ "closureDigest": "sha256:6866f06f64152df2bf378c2788d187cfc1718a83e7471104342ef6818d5da5e7"
}
],
- "digest": "sha256:da78b2c820e608852e318e7828eab47c705e1f15b3657fa879e4316aa9d69b7e"
+ "digest": "sha256:7006af694856f3e7e10e2c5bc00d653fbe7f7702f6172ec292ccbe4de459293f"
}
diff --git a/examples/talking-head-aroll/README.md b/examples/talking-head-aroll/README.md
index 37d471a9..11fc0345 100644
--- a/examples/talking-head-aroll/README.md
+++ b/examples/talking-head-aroll/README.md
@@ -11,10 +11,9 @@ Supply these ignored local assets before checking or building:
Every take uses the same voice-timbre reference. The visual reference sequence is base, alternate,
alternate, base.
-`@narratage/seedance-speaker` binds the reusable project SVS Recipe, Script dialogue and explicit
-references into Text Bindings. The separately imported `official-ugc-v1.svs` compiles to a
-domain-neutral Text Template. A visible Text render Operation produces each prompt and connects it
-to the exact Seedance request; prompt assembly is therefore part of the graph, not hidden package logic.
+The data-only `seedance-kits/speaker` template and project SVS Recipe feed generic Text rendering.
+Each resulting Text output connects to an exact `seedance:ReferenceVideo` beside explicit image,
+audio and duration edges; prompt assembly is therefore part of the graph, not hidden Speaker logic.
The four generated videos are normalized and concatenated into one Speech Spine. One canonical
audio projection goes through WhisperX, direct Script alignment produces the complete SemanticMap,
Gemini plans Cue cuts without seeing timing, and the resulting Caption
@@ -44,7 +43,6 @@ pnpm narratage lock-packages examples/talking-head-aroll/svml.packages.lock \
--package @narratage/text \
--package @narratage/generation \
--package @narratage/seedance \
- --package @narratage/seedance-speaker \
--package @narratage/speech-alignment \
--package @narratage/speech-basis \
--package @narratage/speech-spine \
@@ -120,9 +118,9 @@ logical output as `substitute`. Reverse demand therefore removes all four Seedan
keeping media normalization, WhisperX, Gemini planning, Film and rendering reachable. This is a new
Build, not a continuation or automatic cache hit.
-This public fixture is intentionally brand-neutral. It preserves the topology and the explicit
-fresh/reuse Run shapes used by an internal paid acceptance, but does not claim that its source bytes
-or generated Records are the private delivery inputs from that run.
+This brand-neutral fixture demonstrates the same topology with two explicit Run shapes: fresh
+generation and reuse through zero-input Build-Record Candidates. Local source assets and generated
+Records are intentionally not committed.
The package lock lists independent packages. It is not a hidden video bundle, and the video CLI
contains no authoring or Run-language package by default.
diff --git a/examples/talking-head-aroll/main.svml b/examples/talking-head-aroll/main.svml
index b0857341..c71bde11 100644
--- a/examples/talking-head-aroll/main.svml
+++ b/examples/talking-head-aroll/main.svml
@@ -2,9 +2,10 @@
<svml>
<import from="@narratage/script@1"/>
+ <import as="wording" from="@narratage/text@1"/>
<import as="media" from="@narratage/media@1"/>
<import as="estimate" from="@narratage/estimate@1"/>
- <import as="speaker" from="@narratage/seedance-speaker@1"/>
+ <import as="seedance" from="@narratage/seedance@1"/>
<import as="speech" from="@narratage/speech-spine@1"/>
<import as="whisperx" from="@narratage/whisperx@1"/>
<import as="caption" from="@narratage/caption@1"/>
@@ -15,7 +16,7 @@
<import as="film" from="@narratage/film@1"/>
<import as="render" from="@narratage/render-hyperframes@1"/>
<import as="studio" source="./studio.svs"/>
- <import as="ugc" source="../../packages/seedance-speaker/kits/official-ugc-v1.svs"/>
+ <import as="speaker-kit" source="../../packages/seedance-kits/kits/speaker-v1.svs"/>
<script id="story">
<hook>
@@ -40,56 +41,115 @@
<media:Image id="presenter-alt" src="./assets/presenter-alt.png"/>
<media:Audio id="presenter-voice" src="./assets/presenter-voice.mp3"/>
+ <!-- Per-take performance remains ordinary authored Text, separate from spoken words. -->
+ <wording:Value id="hook-action">
+ Begin with urgent direct eye contact, then let the final admission land more quietly.
+ </wording:Value>
+ <wording:Value id="meeting-action">
+ Briefly indicate the recording tool, then return to the lens with visible relief.
+ </wording:Value>
+ <wording:Value id="evidence-action">
+ Build confidence across the explanation and use one compact connecting gesture.
+ </wording:Value>
+ <wording:Value id="payoff-action">
+ Deliver the reveal with a contained smile, then finish warmly and inclusively.
+ </wording:Value>
+
<!-- Duration remains an explicit graph fact derived from the pronunciation projection. -->
<estimate:Speech id="hook-duration" source={story.segment.hook.speech} policy={studio.speech.normal}/>
<estimate:Speech id="meeting-duration" source={story.segment.meeting.speech} policy={studio.speech.normal}/>
<estimate:Speech id="evidence-duration" source={story.segment.evidence.speech} policy={studio.speech.normal}/>
<estimate:Speech id="payoff-duration" source={story.segment.payoff.speech} policy={studio.speech.normal}/>
+ <!-- Prompt assembly is an ordinary Text subgraph. -->
+ <wording:Render
+ id="hook-prompt"
+ template={speaker-kit.speaker-v1}
+ recipe={studio.speaker.host}
+ >
+ <wording:Set name="dialogue" text={story.segment.hook.dialogue}/>
+ <wording:Set name="action" text={hook-action}/>
+ </wording:Render>
+
+ <wording:Render
+ id="meeting-prompt"
+ template={speaker-kit.speaker-v1}
+ recipe={studio.speaker.host}
+ >
+ <wording:Set name="dialogue" text={story.segment.meeting.dialogue}/>
+ <wording:Set name="action" text={meeting-action}/>
+ </wording:Render>
+
+ <wording:Render
+ id="evidence-prompt"
+ template={speaker-kit.speaker-v1}
+ recipe={studio.speaker.host}
+ >
+ <wording:Set name="dialogue" text={story.segment.evidence.dialogue}/>
+ <wording:Set name="action" text={evidence-action}/>
+ </wording:Render>
+
+ <wording:Render
+ id="payoff-prompt"
+ template={speaker-kit.speaker-v1}
+ recipe={studio.speaker.host}
+ >
+ <wording:Set name="dialogue" text={story.segment.payoff.dialogue}/>
+ <wording:Set name="action" text={payoff-action}/>
+ </wording:Render>
+
<!-- Base → alternate → alternate → base; all four share the same voice reference. -->
- <speaker:Take
+ <seedance:ReferenceVideo
id="hook-take"
- dialogue={story.segment.hook.dialogue}
+ model="mini"
+ prompt={hook-prompt}
duration={hook-duration.duration}
- recipe={studio.speaker.host}
- kit={ugc.official-ugc-v1}
+ resolution="720p"
+ aspect-ratio="9:16"
+ generate-audio="true"
>
- <speaker:Reference image={presenter-clean} role="character-and-scene"/>
- <speaker:Reference audio={presenter-voice} role="voice-timbre"/>
- </speaker:Take>
+ <seedance:Reference image={presenter-clean}/>
+ <seedance:Reference audio={presenter-voice}/>
+ </seedance:ReferenceVideo>
- <speaker:Take
+ <seedance:ReferenceVideo
id="meeting-take"
- dialogue={story.segment.meeting.dialogue}
+ model="mini"
+ prompt={meeting-prompt}
duration={meeting-duration.duration}
- recipe={studio.speaker.host}
- kit={ugc.official-ugc-v1}
+ resolution="720p"
+ aspect-ratio="9:16"
+ generate-audio="true"
>
- <speaker:Reference image={presenter-alt} role="character-and-scene"/>
- <speaker:Reference audio={presenter-voice} role="voice-timbre"/>
- </speaker:Take>
+ <seedance:Reference image={presenter-alt}/>
+ <seedance:Reference audio={presenter-voice}/>
+ </seedance:ReferenceVideo>
- <speaker:Take
+ <seedance:ReferenceVideo
id="evidence-take"
- dialogue={story.segment.evidence.dialogue}
+ model="mini"
+ prompt={evidence-prompt}
duration={evidence-duration.duration}
- recipe={studio.speaker.host}
- kit={ugc.official-ugc-v1}
+ resolution="720p"
+ aspect-ratio="9:16"
+ generate-audio="true"
>
- <speaker:Reference image={presenter-alt} role="character-and-scene"/>
- <speaker:Reference audio={presenter-voice} role="voice-timbre"/>
- </speaker:Take>
+ <seedance:Reference image={presenter-alt}/>
+ <seedance:Reference audio={presenter-voice}/>
+ </seedance:ReferenceVideo>
- <speaker:Take
+ <seedance:ReferenceVideo
id="payoff-take"
- dialogue={story.segment.payoff.dialogue}
+ model="mini"
+ prompt={payoff-prompt}
duration={payoff-duration.duration}
- recipe={studio.speaker.host}
- kit={ugc.official-ugc-v1}
+ resolution="720p"
+ aspect-ratio="9:16"
+ generate-audio="true"
>
- <speaker:Reference image={presenter-clean} role="character-and-scene"/>
- <speaker:Reference audio={presenter-voice} role="voice-timbre"/>
- </speaker:Take>
+ <seedance:Reference image={presenter-clean}/>
+ <seedance:Reference audio={presenter-voice}/>
+ </seedance:ReferenceVideo>
<space:Canvas id="vertical" width="720" height="1280"/>
diff --git a/examples/talking-head-aroll/studio.svs b/examples/talking-head-aroll/studio.svs
index 405086c2..f990ba95 100644
--- a/examples/talking-head-aroll/studio.svs
+++ b/examples/talking-head-aroll/studio.svs
@@ -10,17 +10,11 @@
}
speaker.host {
- kind: ugc-talking-head;
- model: mini;
- resolution: 720p;
- aspect-ratio: 9:16;
-
composition-stability: soft-locked;
camera-motion: none;
edit-rhythm: continuous-take;
performance: natural-explainer;
gesture: natural;
- voice-mode: single-speaker;
}
film.vertical {
diff --git a/examples/talking-head-aroll/svml.packages.lock b/examples/talking-head-aroll/svml.packages.lock
index 3b2a13d7..cd643cf3 100644
--- a/examples/talking-head-aroll/svml.packages.lock
+++ b/examples/talking-head-aroll/svml.packages.lock
@@ -599,7 +599,7 @@
{
"name": "@narratage/fonts-open",
"version": "0.0.0-dev",
- "digest": "sha256:2636ba4fa9323b3d5715db76f819ed702ff55fbc8a1e7f44f0401389f2a7bdb6"
+ "digest": "sha256:14f53b6b08cfa1a24d0e62c80755e9e24891de9fec7285bfca449019fd0cf965"
},
{
"name": "@narratage/generation",
@@ -659,7 +659,7 @@
{
"name": "@narratage/render-hyperframes",
"version": "0.0.0-dev",
- "digest": "sha256:18d5f3a63c71135771af5e9b47f8bcd6cf41662adcb3f1d1628dce6138f75931"
+ "digest": "sha256:5f30fd4c9ce4de8984408d125498f58b06899011d6e07b43bea6ff3498490fe1"
},
{
"name": "@narratage/run-markup",
@@ -676,11 +676,6 @@
"version": "0.0.0-dev",
"digest": "sha256:425cb50348d40a7d14e4bac5d28f32be2f50c9a0504341ad8d72112f6e776d16"
},
- {
- "name": "@narratage/seedance-speaker",
- "version": "0.0.0-dev",
- "digest": "sha256:2b560523007d5781ce4403a4cc8118db094f618fc579043930dcd12c4eb33b71"
- },
{
"name": "@narratage/seedance",
"version": "0.0.0-dev",
@@ -754,7 +749,7 @@
{
"name": "@narratage/whisperx",
"version": "0.0.0-dev",
- "digest": "sha256:7209d52d61e9c9237b59695a30f2f04a055d83053f06439ba9e9a31dd6cd189c"
+ "digest": "sha256:7ee7f559fdc0508b6a7447838eb0a121cbe60c03522c6969638e0e8d3c04068f"
},
{
"name": "cmu-pronouncing-dictionary",
@@ -833,7 +828,7 @@
"version": "0.0.0-dev"
},
"facetsDigest": "sha256:31b51352edcb4f34beccc9d9ac8cf7397125a840971a72c912df3d34d1f3feda",
- "closureDigest": "sha256:cf404e70036afe899420d8271a0fa29e640fbe727fe25020c0a852c33db25cae"
+ "closureDigest": "sha256:31d17d4d2f19841ed32f85ac7f9ffe6e77b45b2514cd4cd52a96a979cc848ba0"
},
{
"specifier": "@narratage/generation",
@@ -896,7 +891,7 @@
"version": "0.0.0-dev"
},
"facetsDigest": "sha256:4210736adc699b3b5f86bf8b0cd84cd7a3c946754c8f19ceae5e0390a3d1d9c5",
- "closureDigest": "sha256:24acdfad8a63da6d5b9f3205e0bd04cb060fafea67b9c0d20623f108514644f0"
+ "closureDigest": "sha256:396057461f2715ea0a2890418e237c35a51ccc379ead2cef85392cdfb54015dd"
},
{
"specifier": "@narratage/run-markup",
@@ -925,15 +920,6 @@
"facetsDigest": "sha256:e889ca3808a19f544e7aa5e27453beb5a09464c645633b0b1b1d9a4b7ce6591a",
"closureDigest": "sha256:29c64c13377e07b6fb100a57c66cda66a05cb671e2d45c663b9bd0c9c926ffac"
},
- {
- "specifier": "@narratage/seedance-speaker",
- "package": {
- "name": "@narratage/seedance-speaker",
- "version": "0.0.0-dev"
- },
- "facetsDigest": "sha256:435c521a965a6b48f3b0821e83da7890b3087535ef8314f3da2515d9f26a71da",
- "closureDigest": "sha256:010c5b9e2d65d0242754ba4df2bc49e842e85d23092397971393aa42b5191cd1"
- },
{
"specifier": "@narratage/semantic-map",
"package": {
@@ -1031,8 +1017,8 @@
"version": "0.0.0-dev"
},
"facetsDigest": "sha256:73fceef213e8d6bcad92532c5d712842e62921743b81bc7bb4d9022a92067359",
- "closureDigest": "sha256:e24bf6745248f12fd2fe968f693e7629800a61a6eafda464dd8c3b08c93215f5"
+ "closureDigest": "sha256:6866f06f64152df2bf378c2788d187cfc1718a83e7471104342ef6818d5da5e7"
}
],
- "digest": "sha256:1e7358b0dd76eb8ba8ed9b9b21f3044e10aa303c4d607cf10db785a7e4805e01"
+ "digest": "sha256:092bff04b86d49a97af233935b64275981add795d1c4ab55624a61ccf4be57c1"
}
diff --git a/package.json b/package.json
index 369b9c65..74184b0b 100644
--- a/package.json
+++ b/package.json
@@ -63,7 +63,6 @@
"@narratage/screen-overlay": "workspace:*",
"@narratage/seedance": "workspace:*",
"@narratage/seedance-kits": "workspace:*",
- "@narratage/seedance-speaker": "workspace:*",
"@narratage/semantic-map": "workspace:*",
"@narratage/speech": "workspace:*",
"@narratage/speech-alignment": "workspace:*",
diff --git a/packages/artifact-store-s3/README.md b/packages/artifact-store-s3/README.md
index 4bd726ae..1a242bfb 100644
--- a/packages/artifact-store-s3/README.md
+++ b/packages/artifact-store-s3/README.md
@@ -16,7 +16,7 @@ behavior in this adapter.
```ts
const artifacts = createS3ArtifactStorePackage({
instance: "artifacts.team",
- bucket: "hypit-svml-artifacts",
+ bucket: "team-svml-artifacts",
prefix: "development",
region: "us-east-1",
expectedBucketOwner: "123456789012",
diff --git a/packages/fonts-open/README.md b/packages/fonts-open/README.md
index 69d7d7e7..42c901cb 100644
--- a/packages/fonts-open/README.md
+++ b/packages/fonts-open/README.md
@@ -72,6 +72,6 @@ families preserve every Unicode-range shard from their installed CSS as one logi
faces referenced by the Author Graph enter a Build transfer bundle and ArtifactStore.
Every catalog entry is either SIL OFL 1.1 or Apache 2.0. Tests compare the declared license against
-the installed package's own metadata. The four private/local Twinit fonts were intentionally not
+the installed package's own metadata. The four private/local legacy fonts were intentionally not
migrated: public redistribution requires an independently verified license, and brand fonts remain
explicit `<media:Font>` author assets.
diff --git a/packages/fonts-open/test/fonts-open.test.ts b/packages/fonts-open/test/fonts-open.test.ts
index 7d58eaea..ee9d793a 100644
--- a/packages/fonts-open/test/fonts-open.test.ts
+++ b/packages/fonts-open/test/fonts-open.test.ts
@@ -66,7 +66,7 @@ test("the catalog exposes 109 publishable open families with frontend metadata",
}
});
-test("every open font from the Twinit catalog is represented without private local fonts", () => {
+test("every redistributable legacy font is represented without private local fonts", () => {
const migrated = [
"caveat", "permanent-marker", "shadows-into-light", "shantell-sans", "playpen-sans",
"short-stack", "kalam", "patrick-hand", "handlee", "coming-soon", "architects-daughter",
diff --git a/packages/image-transform/README.md b/packages/image-transform/README.md
index 7c718d48..04751d19 100644
--- a/packages/image-transform/README.md
+++ b/packages/image-transform/README.md
@@ -10,7 +10,7 @@ The result is the transformed image itself. It contains no source digest, Provid
post-processing report or copied upstream metadata. Core's Derivation records the input edges and
the Need/Receipt records the selected execution endpoint.
-`gptImageDenoiseV1` preserves the useful Twinit GPT Image cleanup exactly: YCrCb NLM with luma 2,
+`gptImageDenoiseV1` preserves the useful legacy GPT Image cleanup exactly: YCrCb NLM with luma 2,
chroma 10, 7/21 windows and 1.02 saturation recovery, followed by PNG encoding. It is now an
explicit reusable Program rather than hidden behavior inside GPT Image generation.
diff --git a/packages/image-transform/test/image-transform.test.ts b/packages/image-transform/test/image-transform.test.ts
index 4cbb5c29..2dafb531 100644
--- a/packages/image-transform/test/image-transform.test.ts
+++ b/packages/image-transform/test/image-transform.test.ts
@@ -15,7 +15,7 @@ import { rasterCapabilities } from "@narratage/raster";
import { digestOf } from "@narratage/protocol";
import type { CanonicalValue, TypedRecord } from "@narratage/protocol";
-test("the Twinit GPT Image cleanup is one explicit reusable Program", () => {
+test("the legacy GPT Image cleanup is one explicit reusable Program", () => {
assert.deepEqual(gptImageDenoiseV1.operations, [{
kind: "denoise",
method: "nlm-ycrcb",
diff --git a/packages/local/README.md b/packages/local/README.md
index fec832dc..79f40122 100644
--- a/packages/local/README.md
+++ b/packages/local/README.md
@@ -63,7 +63,7 @@ export default await createProjectLocalRuntime({
root: import.meta.dirname,
packageLock: "./svml.packages.lock",
runtimeServices: [createS3ArtifactStorePackage({
- bucket: "hypit-svml-artifacts",
+ bucket: "team-svml-artifacts",
prefix: "development",
region: "us-east-1",
})],
diff --git a/packages/local/src/runtime.ts b/packages/local/src/runtime.ts
index ac9aecb3..cc23db20 100644
--- a/packages/local/src/runtime.ts
+++ b/packages/local/src/runtime.ts
@@ -305,7 +305,7 @@ export async function createLocalRuntime(
/**
* Zero-service local distribution. The Runtime authority stays in this process while configured
- * Capability Endpoints may execute locally, in a vendor API, in Lambda, or on Hypit.
+ * Capability Endpoints may execute locally, in a vendor API, in Lambda, or on a hosted service.
*/
export async function createProjectLocalRuntime(
options: ProjectLocalRuntimeOptions = {},
diff --git a/packages/provider-google-vertex/README.md b/packages/provider-google-vertex/README.md
index fc01705e..2089d99e 100644
--- a/packages/provider-google-vertex/README.md
+++ b/packages/provider-google-vertex/README.md
@@ -20,6 +20,6 @@ keeps filesystem authority out of the Provider. A deployment may bind another `C
and pass its `CredentialRef`. Credentials never enter author source, Runtime Closure, BuildState,
Operation state, Prompt or Receipt metadata.
-The old `twinit` implementation used `GOOGLE_APPLICATION_CREDENTIALS` as an ADC file path. A host
+The legacy implementation used `GOOGLE_APPLICATION_CREDENTIALS` as an ADC file path. A host
may preserve that deployment mechanism in a file-backed CredentialStore, but the Provider itself
does not read arbitrary paths or ambient environment variables.
diff --git a/packages/render-hyperframes/README.md b/packages/render-hyperframes/README.md
index 02d5b5de..3ca15bf1 100644
--- a/packages/render-hyperframes/README.md
+++ b/packages/render-hyperframes/README.md
@@ -12,7 +12,7 @@ The package owns `<render:Video composition={...}/>` and lowers it to six ordina
6. expose the verified mux Artifact unchanged as a domain-neutral `BlobArtifact`.
The package is not a renderer and contains no queue, credentials or deployment choice. A local
-HyperFrames process, a hosted Hypit endpoint or another conforming execution package may register a
+HyperFrames process, a hosted endpoint or another conforming execution package may register a
Provider for the exact visual capability. Media Providers independently realize audio rendering and
mux. None parses SVML or decides which Composition to render. Every result is bound to the same
ProgramSpace and exact frame/sample domain before it can become a final video Record.
diff --git a/packages/runtime/test/operations.test.ts b/packages/runtime/test/operations.test.ts
index 516a1f66..c6b1b052 100644
--- a/packages/runtime/test/operations.test.ts
+++ b/packages/runtime/test/operations.test.ts
@@ -24,7 +24,7 @@ test("Operation identity locks Build, Command, Endpoint, implementation and Runt
const first = identity();
assert.equal(first.id, identity().id);
assert.equal(first.submissionKey, identity().submissionKey);
- assert.notEqual(first.id, identity({ endpoint: "hyperframes.hypit" }).id);
+ assert.notEqual(first.id, identity({ endpoint: "hyperframes.hosted" }).id);
assert.notEqual(first.id, identity({ runtimeClosure: digestOf("runtime-closure:hosted") }).id);
});
diff --git a/packages/seedance-kits/README.md b/packages/seedance-kits/README.md
index 72741ca7..03a72719 100644
--- a/packages/seedance-kits/README.md
+++ b/packages/seedance-kits/README.md
@@ -23,8 +23,30 @@ ordinary explicit graph edges:
</seedance:ReferenceVideo>
```
-The six templates are:
+Speaker uses the same graph vocabulary. The Kit assumes `@image1` is the visible person and
+`@audio1` is the voice-timbre reference; it owns no media counting or generation wrapper:
+```svml
+<import as="speaker-kit" source="../../packages/seedance-kits/kits/speaker-v1.svs"/>
+
+<text:Render id="hook-prompt"
+ template={speaker-kit.speaker-v1}
+ recipe={studio.speaker.host}>
+ <text:Set name="dialogue" text={story.segment.hook.dialogue}/>
+ <text:Set name="action" text={hook-action}/>
+</text:Render>
+
+<seedance:ReferenceVideo id="hook-take" model="mini"
+ prompt={hook-prompt} duration={hook-duration.duration}
+ resolution="720p" aspect-ratio="9:16" generate-audio="true">
+ <seedance:Reference image={presenter}/>
+ <seedance:Reference audio={voice}/>
+</seedance:ReferenceVideo>
+```
+
+The seven templates are:
+
+- `speaker-v1`: one visible speaker, one character-and-scene image and one voice reference;
- `broll-v1`: silent visual micro-story;
- `podcast-v1`: two fixed podcast views with two voices;
- `call-v1`: two live video-call reverse views;
@@ -32,6 +54,6 @@ The six templates are:
- `motion-reference-v1`: preserve the subject and transfer body motion only;
- `camera-reference-v1`: preserve the subject and transfer camera language only.
-The first four retain the useful orthogonal prompt axes from Twinit. Historical aliases that emitted
+Speaker, B-roll, Podcast, Call and Street Interview retain useful orthogonal prompt axes from the legacy implementation. Historical aliases that emitted
identical prose were removed. Speed, trim, last-frame extraction and audio extraction are ordinary
media operations and intentionally do not live in these templates.
diff --git a/packages/seedance-kits/kits/broll-v1.svs b/packages/seedance-kits/kits/broll-v1.svs
index 4e222368..c67ad3a6 100644
--- a/packages/seedance-kits/kits/broll-v1.svs
+++ b/packages/seedance-kits/kits/broll-v1.svs
@@ -4,7 +4,6 @@
text-template.broll-v1 {
separator: paragraph;
default-material-mode: practical-real;
- default-reference-plan: single-stage;
default-story-shape: single-moment;
default-edit-language: continuous-shot;
default-camera-language: reference-locked;
@@ -13,7 +12,7 @@
text-template.broll-v1.block.base { kind: fixed; order: 10; text: "Create a silent B-roll micro-film using the provided reference image or images."; }
text-template.broll-v1.block.contract { kind: fixed; order: 20; text: "B-ROLL CONTRACT: create visual support footage. Focus on visible action, object behavior, camera movement, and editing rhythm. Visible people perform visual actions rather than speaking to camera."; }
- text-template.broll-v1.block.reference { kind: fixed; order: 30; text: "REFERENCE CONTRACT: preserve the intended identity, product, UI, props, setting, lighting, texture, wardrobe, and composition from the reference images and story."; }
+ text-template.broll-v1.block.reference { kind: fixed; order: 30; text: "REFERENCE CONTRACT: preserve the intended identity, product, UI, props, setting, lighting, texture, wardrobe, and composition from the reference images. Use the references in their authored order and follow the story's explicit description of each reference's role."; }
text-template.broll-v1.block.material { kind: axis; order: 40; parameter: material-mode; }
text-template.broll-v1.choice.material.practical-real { text: "MATERIAL MODE: practical real-world UGC footage with believable physical causality and a lived-in phone-camera feeling."; }
@@ -22,11 +21,6 @@
text-template.broll-v1.choice.material.proof-capture { text: "MATERIAL MODE: proof capture such as a website, certificate, receipt, review, chart, or credibility page, treated as visual evidence."; }
text-template.broll-v1.choice.material.stylized-vfx { text: "MATERIAL MODE: an intentional visual-effects metaphor used as an expressive explainer rather than practical documentation."; }
- text-template.broll-v1.block.references { kind: axis; order: 50; parameter: reference-plan; }
- text-template.broll-v1.choice.references.single-stage { text: "REFERENCE PLAN: use @image1 as the scene and stage reference for the whole clip."; }
- text-template.broll-v1.choice.references.scene-plus-refs { text: "REFERENCE PLAN: use @image1 as the scene and stage; later images preserve product, UI, logo, prop, or proof details."; }
- text-template.broll-v1.choice.references.ordered-montage { text: "REFERENCE PLAN: use @image1, @image2, @image3 and later images as ordered shots in one same-theme montage."; }
-
text-template.broll-v1.block.story-shape { kind: axis; order: 60; parameter: story-shape; }
text-template.broll-v1.choice.story-shape.single-moment { text: "STORY SHAPE: one continuous visual moment with a small readable change or payoff."; }
text-template.broll-v1.choice.story-shape.reaction-beat { text: "STORY SHAPE: stimulus, realization, then a clearly visible reaction."; }
@@ -55,5 +49,4 @@
text-template.broll-v1.block.hygiene { kind: fixed; order: 100; text: "SCREEN AND TEXT HYGIENE: add no subtitles, captions, floating words, labels, logos, or ad overlays. Referenced physical labels, screens, websites, and documents remain attached to their objects."; }
text-template.broll-v1.block.story { kind: slot; order: 110; slot: story; optional: false; label: "MICRO-STORY PROMPT:"; }
- text-template.broll-v1.block.extra { kind: slot; order: 120; slot: extra; optional: true; }
</sheet>
diff --git a/packages/seedance-kits/kits/call-v1.svs b/packages/seedance-kits/kits/call-v1.svs
index 208c24b5..392357ca 100644
--- a/packages/seedance-kits/kits/call-v1.svs
+++ b/packages/seedance-kits/kits/call-v1.svs
@@ -51,6 +51,5 @@
text-template.call-v1.block.hygiene { kind: fixed; order: 120; text: "AUDIO AND TEXTURE: clean dialogue and realistic screen-recorded call footage. Add no subtitles, captions, chat bubbles, notification cards, or readable overlays; preserve interface elements already present in the references."; }
text-template.call-v1.block.dialogue { kind: slot; order: 130; slot: dialogue; optional: false; label: "DIALOGUE SCRIPT TO DELIVER:"; }
- text-template.call-v1.block.action { kind: slot; order: 140; slot: action; optional: true; }
- text-template.call-v1.block.extra { kind: slot; order: 150; slot: extra; optional: true; }
+ text-template.call-v1.block.action { kind: slot; order: 140; slot: action; optional: true; label: "ACTION DIRECTION:"; }
</sheet>
diff --git a/packages/seedance-kits/kits/camera-reference-v1.svs b/packages/seedance-kits/kits/camera-reference-v1.svs
index 11745216..20386ac3 100644
--- a/packages/seedance-kits/kits/camera-reference-v1.svs
+++ b/packages/seedance-kits/kits/camera-reference-v1.svs
@@ -5,5 +5,5 @@
text-template.camera-reference-v1.block.subject { kind: fixed; order: 10; text: "Use @image1 as the subject and preserve the subject's identity, outfit, and physical features."; }
text-template.camera-reference-v1.block.transfer { kind: fixed; order: 20; text: "Replicate @video1's camera framing, lens choice, camera height, movement, and photographic rhythm, applying that camera language to the subject from @image1 in a coherent setting."; }
text-template.camera-reference-v1.block.exclude { kind: fixed; order: 30; text: "Do not copy @video1's subject pose or body action. Let the subject move naturally for the transferred camera treatment."; }
- text-template.camera-reference-v1.block.extra { kind: slot; order: 40; slot: extra; optional: true; }
+ text-template.camera-reference-v1.block.direction { kind: slot; order: 40; slot: direction; optional: true; label: "ADDITIONAL DIRECTION:"; }
</sheet>
diff --git a/packages/seedance-kits/kits/motion-reference-v1.svs b/packages/seedance-kits/kits/motion-reference-v1.svs
index a433eddc..d61ea644 100644
--- a/packages/seedance-kits/kits/motion-reference-v1.svs
+++ b/packages/seedance-kits/kits/motion-reference-v1.svs
@@ -5,5 +5,5 @@
text-template.motion-reference-v1.block.subject { kind: fixed; order: 10; text: "Use @image1 as the subject and preserve the subject's identity, outfit, and physical features."; }
text-template.motion-reference-v1.block.transfer { kind: fixed; order: 20; text: "Replicate the body motion, gestures, and pose dynamics from @video1 exactly. The subject from @image1 performs that motion in their own setting."; }
text-template.motion-reference-v1.block.exclude { kind: fixed; order: 30; text: "Do not copy @video1's camera framing or lens. Let the camera move naturally to capture the transferred action."; }
- text-template.motion-reference-v1.block.extra { kind: slot; order: 40; slot: extra; optional: true; }
+ text-template.motion-reference-v1.block.direction { kind: slot; order: 40; slot: direction; optional: true; label: "ADDITIONAL DIRECTION:"; }
</sheet>
diff --git a/packages/seedance-kits/kits/podcast-v1.svs b/packages/seedance-kits/kits/podcast-v1.svs
index c62ac0f2..bed639e4 100644
--- a/packages/seedance-kits/kits/podcast-v1.svs
+++ b/packages/seedance-kits/kits/podcast-v1.svs
@@ -51,6 +51,5 @@
text-template.podcast-v1.block.action-contract { kind: fixed; order: 110; text: "ACTION PROMPT: optional segment direction controls specified lines, gestures, expressions, and timing."; }
text-template.podcast-v1.block.audio { kind: fixed; order: 120; text: "AUDIO AND TEXTURE: clean dialogue, natural room tone, stable faces and hands, microphones, props, and photoreal motion. Add no subtitles, captions, labels, UI text, or floating words."; }
text-template.podcast-v1.block.dialogue { kind: slot; order: 130; slot: dialogue; optional: false; label: "DIALOGUE SCRIPT TO DELIVER:"; }
- text-template.podcast-v1.block.action { kind: slot; order: 140; slot: action; optional: true; }
- text-template.podcast-v1.block.extra { kind: slot; order: 150; slot: extra; optional: true; }
+ text-template.podcast-v1.block.action { kind: slot; order: 140; slot: action; optional: true; label: "ACTION DIRECTION:"; }
</sheet>
diff --git a/packages/seedance-kits/kits/speaker-v1.svs b/packages/seedance-kits/kits/speaker-v1.svs
new file mode 100644
index 00000000..c06ca8b1
--- /dev/null
+++ b/packages/seedance-kits/kits/speaker-v1.svs
@@ -0,0 +1,136 @@
+<?svml using="@narratage/text/svs@1"?>
+
+<sheet version="1" id="speaker-v1">
+ text-template.speaker-v1 {
+ separator: paragraph;
+ default-composition-stability: soft-locked;
+ default-camera-motion: none;
+ default-edit-rhythm: continuous-take;
+ default-performance: natural-explainer;
+ default-gesture: natural;
+ }
+
+ text-template.speaker-v1.block.base {
+ kind: fixed;
+ order: 10;
+ text: "Create a realistic vertical phone-shot talking-head video. @image1 is the visible speaker and scene reference. Preserve the person's identity, face, outfit, setting, lighting, lens feel, and broad composition. @audio1 is the speaker's voice-timbre reference.";
+ }
+
+ text-template.speaker-v1.block.script-contract {
+ kind: fixed;
+ order: 20;
+ text: "Deliver the supplied script exactly in order with clear lip-sync. The visible person speaks every line. Do not rewrite, summarize, add, remove, or reorder words.";
+ }
+
+ text-template.speaker-v1.block.composition-stability {
+ kind: axis;
+ order: 30;
+ parameter: composition-stability;
+ }
+
+ text-template.speaker-v1.choice.composition-stability.flexible-ugc {
+ text: "COMPOSITION: preserve the reference person and setup while allowing normal lively creator-video framing variation.";
+ }
+
+ text-template.speaker-v1.choice.composition-stability.soft-locked {
+ text: "COMPOSITION: keep the speaker in the same general region and preserve the reference composition, allowing natural posture changes and tiny crop or scale breathing.";
+ }
+
+ text-template.speaker-v1.choice.composition-stability.strict-locked {
+ text: "COMPOSITION: preserve subject placement, perspective, left-right orientation, background relationship, and framing envelope throughout the shot.";
+ }
+
+ text-template.speaker-v1.block.camera-motion {
+ kind: axis;
+ order: 40;
+ parameter: camera-motion;
+ }
+
+ text-template.speaker-v1.choice.camera-motion.none {
+ text: "CAMERA: hold a steady fixed framing.";
+ }
+
+ text-template.speaker-v1.choice.camera-motion.subtle-punch-in-return {
+ text: "CAMERA: use one subtle push-in on the most important phrase, then return smoothly to the original framing.";
+ }
+
+ text-template.speaker-v1.block.edit-rhythm {
+ kind: axis;
+ order: 50;
+ parameter: edit-rhythm;
+ }
+
+ text-template.speaker-v1.choice.edit-rhythm.continuous-take {
+ text: "EDIT RHYTHM: one continuous creator explanation with natural breath timing.";
+ }
+
+ text-template.speaker-v1.choice.edit-rhythm.pause-trim-jump-cuts {
+ text: "EDIT RHYTHM: polished pause-trim jump cuts at phrase boundaries while keeping the framing coherent.";
+ }
+
+ text-template.speaker-v1.block.performance {
+ kind: axis;
+ order: 60;
+ parameter: performance;
+ }
+
+ text-template.speaker-v1.choice.performance.natural-explainer {
+ text: "PERFORMANCE: natural explainer energy with clear eye focus, expressive eyebrows, small nods, and a friendly conversational rhythm.";
+ }
+
+ text-template.speaker-v1.choice.performance.high-energy-ugc {
+ text: "PERFORMANCE: vivid high-energy short-form delivery with expressive face changes on key words.";
+ }
+
+ text-template.speaker-v1.choice.performance.calm-authority {
+ text: "PERFORMANCE: composed, confident, precise, and trustworthy, with deliberate emphasis.";
+ }
+
+ text-template.speaker-v1.choice.performance.reactive-playful {
+ text: "PERFORMANCE: playful and reactive, with amused looks, quick smiles, skeptical eyebrows, and lively timing.";
+ }
+
+ text-template.speaker-v1.block.gesture {
+ kind: axis;
+ order: 70;
+ parameter: gesture;
+ }
+
+ text-template.speaker-v1.choice.gesture.restrained {
+ text: "GESTURE: restrained body language; let face, eyes, and voice carry the meaning.";
+ }
+
+ text-template.speaker-v1.choice.gesture.compact {
+ text: "GESTURE: compact explanatory hand movement close to the torso and inside the frame.";
+ }
+
+ text-template.speaker-v1.choice.gesture.natural {
+ text: "GESTURE: natural creator gestures such as open palms, small pointing beats, nods, and posture shifts where the script calls for them.";
+ }
+
+ text-template.speaker-v1.choice.gesture.expressive {
+ text: "GESTURE: richer hand, shoulder, and posture changes that support the spoken idea while remaining physically coherent.";
+ }
+
+ text-template.speaker-v1.block.hygiene {
+ kind: fixed;
+ order: 80;
+ text: "Keep face, hands, skin texture, lighting, and motion photoreal and stable. Add no subtitles, captions, labels, logos, UI text, floating words, or other readable on-screen text.";
+ }
+
+ text-template.speaker-v1.block.dialogue {
+ kind: slot;
+ order: 90;
+ slot: dialogue;
+ optional: false;
+ label: "SCRIPT TO DELIVER:";
+ }
+
+ text-template.speaker-v1.block.action {
+ kind: slot;
+ order: 100;
+ slot: action;
+ optional: true;
+ label: "ACTION DIRECTION:";
+ }
+</sheet>
diff --git a/packages/seedance-kits/kits/street-interview-v1.svs b/packages/seedance-kits/kits/street-interview-v1.svs
index cc8bf634..000585a7 100644
--- a/packages/seedance-kits/kits/street-interview-v1.svs
+++ b/packages/seedance-kits/kits/street-interview-v1.svs
@@ -51,6 +51,5 @@
text-template.street-interview-v1.block.hygiene { kind: fixed; order: 120; text: "AUDIO AND TEXTURE: clean dialogue, natural outdoor room tone, stable faces, hands, microphone, and photoreal motion. Add no subtitles, captions, labels, cards, or overlays; preserve physical signs and text already in the scene."; }
text-template.street-interview-v1.block.dialogue { kind: slot; order: 130; slot: dialogue; optional: false; label: "DIALOGUE SCRIPT TO DELIVER:"; }
- text-template.street-interview-v1.block.action { kind: slot; order: 140; slot: action; optional: true; }
- text-template.street-interview-v1.block.extra { kind: slot; order: 150; slot: extra; optional: true; }
+ text-template.street-interview-v1.block.action { kind: slot; order: 140; slot: action; optional: true; label: "ACTION DIRECTION:"; }
</sheet>
diff --git a/packages/seedance-kits/package.json b/packages/seedance-kits/package.json
index 9bac0059..d2ad05a2 100644
--- a/packages/seedance-kits/package.json
+++ b/packages/seedance-kits/package.json
@@ -4,6 +4,7 @@
"private": true,
"type": "module",
"exports": {
+ "./speaker": "./kits/speaker-v1.svs",
"./broll": "./kits/broll-v1.svs",
"./podcast": "./kits/podcast-v1.svs",
"./call": "./kits/call-v1.svs",
diff --git a/packages/seedance-kits/test/kits.test.ts b/packages/seedance-kits/test/kits.test.ts
index 29497401..0c1f0e99 100644
--- a/packages/seedance-kits/test/kits.test.ts
+++ b/packages/seedance-kits/test/kits.test.ts
@@ -5,6 +5,7 @@ import { parseSvs } from "@narratage/svs";
import { renderText, sealTextBindings, textTemplateFromSvsRecipes } from "@narratage/text";
const cases = [
+ { file: "speaker-v1.svs", id: "speaker-v1", bindings: { dialogue: "HOST: Meaning comes first." }, marker: "@audio1 is the speaker's voice-timbre reference" },
{ file: "broll-v1.svs", id: "broll-v1", bindings: { story: "A hand opens the product." }, marker: "silent B-roll" },
{ file: "podcast-v1.svs", id: "podcast-v1", bindings: { dialogue: "A: Hello.\nB: Hi." }, marker: "Host A uses @audio1" },
{ file: "call-v1.svs", id: "call-v1", bindings: { dialogue: "A: Hello.\nB: Hi." }, marker: "both tiles are live feeds" },
@@ -22,3 +23,16 @@ test("Seedance Kits are finite data programs with distinct rendered semantics",
assert.match(output.value, new RegExp(item.marker, "u"), item.file);
}
});
+
+test("official Seedance Kits select semantics by file and explicit axes, never inferred branches", () => {
+ for (const item of cases) {
+ const source = readFileSync(new URL(`../kits/${item.file}`, import.meta.url), "utf8");
+ assert.doesNotMatch(source, /kind:\s*variant|when-(?:param|select)-/u, item.file);
+ assert.doesNotMatch(source, /slot:\s*extra/u, item.file);
+ }
+ const broll = readFileSync(new URL("../kits/broll-v1.svs", import.meta.url), "utf8");
+ assert.doesNotMatch(broll, /reference-plan/u);
+ const speaker = readFileSync(new URL("../kits/speaker-v1.svs", import.meta.url), "utf8");
+ assert.match(speaker, /slot:\s*dialogue/u);
+ assert.match(speaker, /slot:\s*action/u);
+});
diff --git a/packages/seedance-speaker/README.md b/packages/seedance-speaker/README.md
deleted file mode 100644
index 20cfb902..00000000
--- a/packages/seedance-speaker/README.md
+++ /dev/null
@@ -1,69 +0,0 @@
-# `@narratage/seedance-speaker`
-
-Official deterministic UGC talking-head author module for Seedance. It is currently an internal
-private workspace unit, not a commitment to publish one npm package per Seedance author module.
-
-The package consumes an inert project SVS Recipe, one explicitly imported `TextTemplate`, one Script
-dialogue `Text` and explicit image/audio references. Its Surface connects that authored `Text`
-directly to a visible `@narratage/text` render component in the graph. Optional per-Take `action`
-and `extra` values are also ordinary Text edges; they are not
-hidden Recipe strings. The rendered `Text` output is wired to Seedance's exact `prompt` port. The
-same Surface authors the low-level
-`@narratage/seedance` `DurationProgram`; generation reuses the exact Seedance ReferenceVideo request shape. The
-official Text Template source is
-[`official-ugc-v1.svs`](./kits/official-ugc-v1.svs).
-
-It owns no Provider, credentials, queue or HTTP behavior. KIE, Volcengine or another explicitly
-bound Endpoint only receives the already compiled exact Seedance request.
-
-The official `official-ugc-v1` template contains a fixed base, reference contract, script contract,
-texture and visible-text exclusion plus six configurable axes: composition stability, camera
-motion, edit rhythm, performance, gesture and voice mapping. Project SVS selects values. The
-official SVS declares defaults, block order, value-to-text choices and finite conditions. SVS
-itself executes neither conditions nor concatenation; the domain-neutral Text component does that.
-
-## Compilation boundary
-
-```text
-Script dialogue Text ──────────────────────────┐
-optional action/extra Text ────────────────────┼──► render Text
-SVS-selected TextBindings + TextTemplate ──────┘ │
- │ exact prompt edge
-authored Blob references + Seedance DurationProgram ────┼──► generation Need
-```
-
-`take.prompt` and `take.video` are ordinary Logical Outputs; `take.program` is authored configuration.
-Prompt assembly is the deterministic `@narratage/text#render` Operation in the BuildPlan, so the text
-can be targeted or satisfied by another compatible Candidate without a Seedance-specific rule.
-
-The author imports this logical module and imports the selected template as an ordinary Source Module.
-The Host package lock activates `@narratage/text` and low-level `@narratage/seedance` independently. A
-future public distribution may offer an optional install bundle, but the current repository keeps
-their activations separate.
-
-## `official-ugc-v1`
-
-Defaults:
-
-| Setting | Default | Choices |
-|---|---|---|
-| model | `seedance-2-mini` | `mini`, `fast`, `standard` |
-| resolution | `720p` | `480p`, `720p`; `1080p` only for standard |
-| aspect ratio | `9:16` | Seedance-supported aspect ratios |
-| web search | `false` | `true`, `false` |
-| composition stability | `soft-locked` | `flexible-ugc`, `soft-locked`, `strict-locked` |
-| camera motion | `none` | `none`, `subtle-punch-in-return` |
-| edit rhythm | `continuous-take` | `continuous-take`, `pause-trim-jump-cuts` |
-| performance | `natural-explainer` | `natural-explainer`, `high-energy-ugc`, `calm-authority`, `reactive-playful` |
-| gesture | `natural` | `restrained`, `compact`, `natural`, `expressive` |
-| voice mode | `single-speaker` | `single-speaker`, `role-tagged-lines` |
-
-The fixed base requests realistic phone-shot UGC, exact ordered dialogue and visible-speaker
-lip-sync, coherent reference identity, photoreal texture and no generated readable text. It always
-requests generated audio. One to nine image references are required; zero to three audio references
-are allowed. With an audio reference, `@audio1` supplies the visible speaker's voice timbre.
-
-`action` and `extra` are optional per-Take Text inputs appended as separate attributable blocks.
-They do not alter Kit control flow. Recipe files select method and style axes only; putting content
-in `recipe.action` or `recipe.extra` is rejected so content dependencies cannot disappear into
-Frontend lowering.
diff --git a/packages/seedance-speaker/kits/official-ugc-v1.svs b/packages/seedance-speaker/kits/official-ugc-v1.svs
deleted file mode 100644
index 263f9f11..00000000
--- a/packages/seedance-speaker/kits/official-ugc-v1.svs
+++ /dev/null
@@ -1,205 +0,0 @@
-<?svml using="@narratage/text/svs@1"?>
-
-<sheet version="1" id="official-ugc-v1">
- text-template.official-ugc-v1 {
- separator: paragraph;
- default-composition-stability: soft-locked;
- default-camera-motion: none;
- default-edit-rhythm: continuous-take;
- default-performance: natural-explainer;
- default-gesture: natural;
- default-voice-mode: single-speaker;
- }
-
- text-template.official-ugc-v1.block.base {
- kind: fixed;
- order: 10;
- text: "Create a realistic vertical iPhone UGC talking-head video using @image1 as the character-and-scene reference.";
- }
-
- text-template.official-ugc-v1.block.reference-contract {
- kind: variant;
- order: 20;
- }
-
- text-template.official-ugc-v1.choice.reference-contract.one {
- when-select-image-count: one;
- text: "REFERENCE CONTRACT: preserve @image1's visible person, face identity, outfit, room, lighting, lens feel, and broad camera style.";
- }
-
- text-template.official-ugc-v1.choice.reference-contract.many {
- when-select-image-count: many;
- text: "REFERENCE CONTRACT: use @image1 as the primary character-and-scene reference. Use @image2 and later image references only as supplementary identity, wardrobe, product, or environment evidence; preserve one coherent person and scene.";
- }
-
- text-template.official-ugc-v1.block.script-contract {
- kind: fixed;
- order: 30;
- text: "SCRIPT: deliver the scripted lines exactly in order, with clear lip-sync for the visible speaker's spoken lines.";
- }
-
- text-template.official-ugc-v1.block.voice-map {
- kind: variant;
- order: 40;
- }
-
- text-template.official-ugc-v1.choice.voice-map.single-without-audio {
- when-param-voice-mode: single-speaker;
- when-select-audio-count: zero;
- text: "VOICE MAP: the visible person in @image1 delivers the full script on camera with a natural voice matching the character.";
- }
-
- text-template.official-ugc-v1.choice.voice-map.single-with-audio {
- when-param-voice-mode: single-speaker;
- when-select-audio-count: one;
- text: "VOICE MAP: the visible person in @image1 delivers the full script on camera. @audio1 is the voice-timbre reference for the visible speaker.";
- }
-
- text-template.official-ugc-v1.choice.voice-map.single-with-many-audios {
- when-param-voice-mode: single-speaker;
- when-select-audio-count: many;
- text: "VOICE MAP: the visible person in @image1 delivers the full script on camera. @audio1 is the voice-timbre reference for the visible speaker. Additional audio references provide supplementary voice evidence only.";
- }
-
- text-template.official-ugc-v1.choice.voice-map.roles-without-audio {
- when-param-voice-mode: role-tagged-lines;
- when-select-audio-count: zero;
- text: "VOICE MAP: the script uses role labels such as HOST: and NARRATOR:. The visible person in @image1 delivers the first speaking role on camera; lines assigned to the second role play as an off-camera narrator voice while the visible speaker stays present and reacts naturally.";
- }
-
- text-template.official-ugc-v1.choice.voice-map.roles-with-one-audio {
- when-param-voice-mode: role-tagged-lines;
- when-select-audio-count: one;
- text: "VOICE MAP: the script uses role labels such as HOST: and NARRATOR:. The visible person in @image1 delivers the first speaking role on camera; lines assigned to the second role play as an off-camera narrator voice while the visible speaker stays present and reacts naturally. @audio1 is the visible speaker's voice-timbre reference.";
- }
-
- text-template.official-ugc-v1.choice.voice-map.roles-with-many-audios {
- when-param-voice-mode: role-tagged-lines;
- when-select-audio-count: many;
- text: "VOICE MAP: the script uses role labels such as HOST: and NARRATOR:. The visible person in @image1 delivers the first speaking role on camera; lines assigned to the second role play as an off-camera narrator voice while the visible speaker stays present and reacts naturally. @audio1 is the visible speaker's voice-timbre reference. @audio2 is the narrator voice-timbre reference.";
- }
-
- text-template.official-ugc-v1.block.composition-stability {
- kind: axis;
- order: 50;
- parameter: composition-stability;
- }
-
- text-template.official-ugc-v1.choice.composition-stability.flexible-ugc {
- text: "COMPOSITION STABILITY: preserve @image1's visible person and overall reference setup while allowing normal short-form UGC framing variation for a lively creator video.";
- }
-
- text-template.official-ugc-v1.choice.composition-stability.soft-locked {
- text: "COMPOSITION STABILITY: keep the speaker in the same general region and maintain the reference composition, allowing natural posture changes and only tiny crop or scale breathing.";
- }
-
- text-template.official-ugc-v1.choice.composition-stability.strict-locked {
- text: "COMPOSITION STABILITY: treat @image1 as a locked spatial stage. Preserve the same subject placement, perspective, left-right orientation, background relationship, and framing envelope so persistent overlays stay spatially meaningful.";
- }
-
- text-template.official-ugc-v1.block.camera-motion {
- kind: axis;
- order: 60;
- parameter: camera-motion;
- }
-
- text-template.official-ugc-v1.choice.camera-motion.none {
- text: "CAMERA MOTION: hold the framing steady in a fixed shot, keeping the selected composition envelope stable from start to finish.";
- }
-
- text-template.official-ugc-v1.choice.camera-motion.subtle-punch-in-return {
- text: "CAMERA MOTION: use one subtle push-in on the most important phrase along the same optical axis, then return smoothly to the original framing while preserving the same subject placement and background relationship.";
- }
-
- text-template.official-ugc-v1.block.edit-rhythm {
- kind: axis;
- order: 70;
- parameter: edit-rhythm;
- }
-
- text-template.official-ugc-v1.choice.edit-rhythm.continuous-take {
- text: "EDIT RHYTHM: make the clip feel like a continuous creator explanation, with natural breath timing and polished short-form pacing.";
- }
-
- text-template.official-ugc-v1.choice.edit-rhythm.pause-trim-jump-cuts {
- text: "EDIT RHYTHM: use pause-trim jump cuts. The final clip feels tightly edited because silent gaps between spoken phrases have been removed; framing stays locked, with only tiny temporal jumps in expression, mouth position, gaze, hands, and posture at phrase boundaries.";
- }
-
- text-template.official-ugc-v1.block.performance {
- kind: axis;
- order: 80;
- parameter: performance;
- }
-
- text-template.official-ugc-v1.choice.performance.natural-explainer {
- text: "PERFORMANCE: natural explainer energy. The speaker uses clear eye focus, expressive eyebrows, small nods, and a friendly teaching rhythm that matches each idea.";
- }
-
- text-template.official-ugc-v1.choice.performance.high-energy-ugc {
- text: "PERFORMANCE: high-energy short-form UGC delivery. The speaker feels magnetic, quick, vivid, and excited, with expressive face changes on key words.";
- }
-
- text-template.official-ugc-v1.choice.performance.calm-authority {
- text: "PERFORMANCE: calm authority. The speaker is composed, confident, precise, and trustworthy, with deliberate emphasis.";
- }
-
- text-template.official-ugc-v1.choice.performance.reactive-playful {
- text: "PERFORMANCE: reactive and playful. The speaker responds visibly to the idea with amused looks, quick smiles, skeptical eyebrows, and lively conversational timing.";
- }
-
- text-template.official-ugc-v1.block.gesture {
- kind: axis;
- order: 90;
- parameter: gesture;
- }
-
- text-template.official-ugc-v1.choice.gesture.restrained {
- text: "GESTURE: restrained body language. Let face, eye focus, and voice carry most of the meaning, with small posture shifts and very light hand beats.";
- }
-
- text-template.official-ugc-v1.choice.gesture.compact {
- text: "GESTURE: compact explanatory gestures. Keep hand movement close to the torso and inside the frame, using small semantic beats that support the words without competing with persistent overlays.";
- }
-
- text-template.official-ugc-v1.choice.gesture.natural {
- text: "GESTURE: natural creator gestures. Use open palms, small pointing beats, nods, and posture shifts where the script naturally calls for them.";
- }
-
- text-template.official-ugc-v1.choice.gesture.expressive {
- text: "GESTURE: expressive embodied delivery. Use richer hand, shoulder, and posture changes that clearly visualize the idea while staying coherent with the spoken line.";
- }
-
- text-template.official-ugc-v1.block.texture {
- kind: fixed;
- order: 100;
- text: "TEXTURE: crisp real-shot phone footage with stable face, hands, skin texture, lighting, and photoreal motion.";
- }
-
- text-template.official-ugc-v1.block.visible-text {
- kind: fixed;
- order: 110;
- text: "No subtitles, captions, labels, logos, UI text, floating words, or any readable on-screen text inside the generated video.";
- }
-
- text-template.official-ugc-v1.block.dialogue {
- kind: slot;
- order: 120;
- slot: dialogue;
- optional: false;
- label: "SCRIPT TO DELIVER:";
- }
-
- text-template.official-ugc-v1.block.action {
- kind: slot;
- order: 130;
- slot: action;
- optional: true;
- }
-
- text-template.official-ugc-v1.block.extra {
- kind: slot;
- order: 140;
- slot: extra;
- optional: true;
- }
-</sheet>
diff --git a/packages/seedance-speaker/package.json b/packages/seedance-speaker/package.json
deleted file mode 100644
index 5269418f..00000000
--- a/packages/seedance-speaker/package.json
+++ /dev/null
@@ -1,24 +0,0 @@
-{
- "name": "@narratage/seedance-speaker",
- "version": "0.0.0-dev",
- "private": true,
- "type": "module",
- "exports": {
- ".": "./src/index.ts"
- },
- "svml": {
- "activation": "./src/activation.ts"
- },
- "dependencies": {
- "@narratage/artifact": "workspace:*",
- "@narratage/elaborator": "workspace:*",
- "@narratage/generation": "workspace:*",
- "@narratage/model-kit": "workspace:*",
- "@narratage/protocol": "workspace:*",
- "@narratage/seedance": "workspace:*",
- "@narratage/speech": "workspace:*",
- "@narratage/svs": "workspace:*",
- "@narratage/text": "workspace:*",
- "@narratage/markup": "workspace:*"
- }
-}
diff --git a/packages/seedance-speaker/src/activation.ts b/packages/seedance-speaker/src/activation.ts
deleted file mode 100644
index 444e1bfa..00000000
--- a/packages/seedance-speaker/src/activation.ts
+++ /dev/null
@@ -1,26 +0,0 @@
-import { createMarkupSurfaceHostFacet } from "@narratage/markup";
-
-import {
- decodeSeedanceSpeakerTakeSurface,
- seedanceSpeakerImplementationDigests,
- seedanceSpeakerManifest,
- seedanceSpeakerModuleRef,
-} from "./index.js";
-
-export const svmlPackage = {
- format: "svml.node-package@1" as const,
- name: "@narratage/seedance-speaker",
- modules: [{
- manifest: seedanceSpeakerManifest,
- specifiers: ["@narratage/seedance-speaker", "@narratage/seedance-speaker@1"],
- }],
- hostFacets: [createMarkupSurfaceHostFacet({
- module: seedanceSpeakerModuleRef,
- surface: "take",
- mode: "structured",
- implementationDigest: seedanceSpeakerImplementationDigests.takeSurface,
- handler: decodeSeedanceSpeakerTakeSurface,
- })],
-};
-
-export default svmlPackage;
diff --git a/packages/seedance-speaker/src/fragment.ts b/packages/seedance-speaker/src/fragment.ts
deleted file mode 100644
index 0f25c631..00000000
--- a/packages/seedance-speaker/src/fragment.ts
+++ /dev/null
@@ -1,21 +0,0 @@
-import type { ExactModelEndpoint, ExactModelMediaInput } from "@narratage/model-kit";
-import type { ProducerRef } from "@narratage/protocol";
-import { createSeedanceDurationGenerationFragment } from "@narratage/seedance";
-
-/**
- * Speaker is an authoring Kit, not another execution protocol. Once its prompt
- * program is compiled, it uses Seedance's ordinary graph-native request
- * assembly, including every runtime-produced media edge.
- */
-export function createSeedanceSpeakerTakeFragment(
- endpoint: ExactModelEndpoint,
- compileRequestProducer: ProducerRef,
- mediaInputs: readonly ExactModelMediaInput[] = [],
-) {
- return createSeedanceDurationGenerationFragment(
- endpoint,
- compileRequestProducer,
- mediaInputs,
- [{ name: "prompt", port: "prompt" }],
- );
-}
diff --git a/packages/seedance-speaker/src/index.ts b/packages/seedance-speaker/src/index.ts
deleted file mode 100644
index 44333178..00000000
--- a/packages/seedance-speaker/src/index.ts
+++ /dev/null
@@ -1,16 +0,0 @@
-export { createSeedanceSpeakerTakeFragment } from "./fragment.js";
-export {
- createSpeakerDurationProgram,
- createSpeakerTextBindings,
- sealSpeakerTakeIntent,
- speakerMethodDefaults,
- verifySpeakerTakeIntent,
-} from "./kit.js";
-export {
- seedanceSpeakerImplementationDigests,
- seedanceSpeakerManifest,
- seedanceSpeakerManifestDigest,
- seedanceSpeakerModuleRef,
-} from "./manifest.js";
-export { decodeSeedanceSpeakerTakeSurface } from "./surface.js";
-export type * from "./types.js";
diff --git a/packages/seedance-speaker/src/kit.ts b/packages/seedance-speaker/src/kit.ts
deleted file mode 100644
index bff1f993..00000000
--- a/packages/seedance-speaker/src/kit.ts
+++ /dev/null
@@ -1,99 +0,0 @@
-import { sealTextBindings } from "@narratage/text";
-import type { TextBindings, TextScalar } from "@narratage/text";
-import {
- sealSeedanceDurationProgram,
- seedanceModels,
-} from "@narratage/seedance";
-import type { SeedanceDurationProgram } from "@narratage/seedance";
-
-import type { SpeakerTakeIntent } from "./types.js";
-
-export const speakerMethodDefaults = {
- model: "seedance-2-mini",
- resolution: "720p",
- aspectRatio: "9:16",
- webSearch: false,
-} as const;
-
-export function sealSpeakerTakeIntent(value: SpeakerTakeIntent): SpeakerTakeIntent {
- const intent = structuredClone(value);
- verifySpeakerTakeIntent(intent);
- return intent;
-}
-
-export function verifySpeakerTakeIntent(value: unknown): asserts value is SpeakerTakeIntent {
- if (value === null || typeof value !== "object" || Array.isArray(value)) {
- throw new Error("Speaker TakeIntent must be an object");
- }
- const intent = value as SpeakerTakeIntent;
- if (
- intent.contract !== "svml.seedance-speaker-take-intent@1"
- || !/^[a-z][a-z0-9-]{0,95}$/u.test(intent.kit)
- || !seedanceModels.includes(intent.model)
- || !["480p", "720p", "1080p"].includes(intent.resolution)
- || !["1:1", "4:3", "3:4", "16:9", "9:16", "21:9", "adaptive"].includes(intent.aspectRatio)
- || typeof intent.webSearch !== "boolean"
- ) {
- throw new Error("Speaker TakeIntent method settings are invalid");
- }
- if (intent.resolution === "1080p" && intent.model !== "seedance-2") {
- throw new Error("Seedance Mini and Fast Speaker Kits support at most 720p");
- }
- if (
- typeof intent.recipe?.path !== "string"
- || intent.recipe.path.length === 0
- || intent.promptParameters === null
- || typeof intent.promptParameters !== "object"
- || Array.isArray(intent.promptParameters)
- ) {
- throw new Error("Speaker TakeIntent identity or parameters are invalid");
- }
- for (const [name, item] of Object.entries(intent.promptParameters)) {
- if (!/^[a-z][a-z0-9-]{0,95}$/u.test(name)) throw new Error(`Speaker prompt parameter ${name} is invalid`);
- if (typeof item !== "string" && typeof item !== "boolean" && !(typeof item === "number" && Number.isFinite(item))) {
- throw new Error(`Speaker prompt parameter ${name} must be a finite scalar`);
- }
- }
- const images = intent.references.filter((item) => item.kind === "image");
- const audios = intent.references.filter((item) => item.kind === "audio");
- if (images.length < 1 || images.length > 9 || audios.length > 3) {
- throw new Error("Speaker requires 1-9 image references and at most 3 audio references");
- }
- for (const item of intent.references) {
- if ((item.kind !== "image" && item.kind !== "audio") || typeof item.role !== "string" || item.role.length === 0) {
- throw new Error("Speaker reference is invalid");
- }
- }
-}
-
-function audioCount(intent: SpeakerTakeIntent): "zero" | "one" | "many" {
- const count = intent.references.filter((item) => item.kind === "audio").length;
- return count === 0 ? "zero" : count === 1 ? "one" : "many";
-}
-
-export function createSpeakerTextBindings(intent: SpeakerTakeIntent): TextBindings {
- verifySpeakerTakeIntent(intent);
- const values: Record<string, TextScalar> = {};
- for (const [name, value] of Object.entries(intent.promptParameters)) {
- values[name] = value as TextScalar;
- }
- return sealTextBindings({
- ...values,
- "image-count": intent.references.filter((item) => item.kind === "image").length === 1 ? "one" : "many",
- "audio-count": audioCount(intent),
- });
-}
-
-export function createSpeakerDurationProgram(intent: SpeakerTakeIntent): SeedanceDurationProgram {
- verifySpeakerTakeIntent(intent);
- return sealSeedanceDurationProgram({
- contract: "svml.seedance-duration-program@1",
- model: intent.model,
- ports: {
- resolution: [intent.resolution],
- aspectRatio: [intent.aspectRatio],
- generateAudio: [true],
- webSearch: [intent.webSearch],
- },
- });
-}
diff --git a/packages/seedance-speaker/src/manifest.ts b/packages/seedance-speaker/src/manifest.ts
deleted file mode 100644
index c08a0102..00000000
--- a/packages/seedance-speaker/src/manifest.ts
+++ /dev/null
@@ -1,56 +0,0 @@
-import { artifactDependency } from "@narratage/artifact";
-import { speechDependency } from "@narratage/speech";
-import { artifactTypes } from "@narratage/artifact";
-import { textDependency, textTypes } from "@narratage/text";
-import { digestOf } from "@narratage/protocol";
-import type { ModuleManifest } from "@narratage/protocol";
-import {
- seedanceManifestDigest,
- seedanceModuleRef,
- seedanceEndpointsByModel,
- seedanceTypes,
-} from "@narratage/seedance";
-import { svsManifest, svsRecipeType } from "@narratage/svs";
-
-export const seedanceSpeakerModuleRef = { name: "@narratage/seedance-speaker", version: "1" } as const;
-
-export const seedanceSpeakerImplementationDigests = {
- takeSurface: digestOf("@narratage/seedance-speaker/take-surface@1"),
-} as const;
-
-export const seedanceSpeakerManifest: ModuleManifest = {
- format: "svml.module@1",
- name: seedanceSpeakerModuleRef.name,
- version: seedanceSpeakerModuleRef.version,
- dependencies: [
- artifactDependency,
- speechDependency,
- textDependency,
- { module: seedanceModuleRef, digest: seedanceManifestDigest },
- { module: svsRecipeType.module, digest: digestOf(svsManifest) },
- ],
- types: [],
- capabilities: [],
- surfaces: [{
- name: "take",
- tag: "Take",
- mode: "structured",
- outputs: [
- textTypes.bindings,
- textTypes.binding,
- textTypes.text,
- seedanceTypes.durationProgram,
- artifactTypes.blob,
- ...Object.values(seedanceEndpointsByModel).flatMap((endpoint) =>
- Object.values(endpoint.mediaBindings).map((binding) => binding.type)),
- ],
- implementation: {
- kind: "trusted-frontend-surface",
- locator: "@narratage/seedance-speaker/take-surface",
- digest: seedanceSpeakerImplementationDigests.takeSurface,
- },
- }],
- producers: [],
-};
-
-export const seedanceSpeakerManifestDigest = digestOf(seedanceSpeakerManifest);
diff --git a/packages/seedance-speaker/src/surface.ts b/packages/seedance-speaker/src/surface.ts
deleted file mode 100644
index ad55e477..00000000
--- a/packages/seedance-speaker/src/surface.ts
+++ /dev/null
@@ -1,377 +0,0 @@
-import { speechTypes } from "@narratage/speech";
-import type { SpeechDuration } from "@narratage/speech";
-import { artifactTypes } from "@narratage/artifact";
-import type { CanonicalValue, StoredValue } from "@narratage/protocol";
-import { generationPort, sealGenerationMediaBinding } from "@narratage/generation";
-import type { GenerationMediaPort } from "@narratage/generation";
-import { exactModelMediaInputNames, exactModelTextInputName } from "@narratage/model-kit";
-import type { ExactModelEndpoint, ExactModelMediaInput } from "@narratage/model-kit";
-import {
- createTextRenderFragment,
- sealTextBinding,
- textTypes,
- verifyText,
- verifyTextTemplate,
-} from "@narratage/text";
-import type { TextTemplate } from "@narratage/text";
-import {
- seedanceEndpointsByModel,
- seedanceDurationCompileProducers,
- seedanceTypes,
-} from "@narratage/seedance";
-import type { SeedanceModel } from "@narratage/seedance";
-import { svsRecipeType } from "@narratage/svs";
-import type { SvsRecipe } from "@narratage/svs";
-import type {
- StructuredElement,
- StructuredSurfaceHandler,
- SurfaceResolvedReference,
- MarkupAttributeValue,
-} from "@narratage/markup";
-
-import { createSeedanceSpeakerTakeFragment } from "./fragment.js";
-import {
- createSpeakerDurationProgram,
- createSpeakerTextBindings,
- sealSpeakerTakeIntent,
- speakerMethodDefaults,
-} from "./kit.js";
-import type { SpeakerReference } from "./types.js";
-
-function localName(value: string): string {
- return value.includes(":") ? value.slice(value.lastIndexOf(":") + 1) : value;
-}
-
-function sameType(left: SurfaceResolvedReference["type"], right: SurfaceResolvedReference["type"]): boolean {
- return left.module.name === right.module.name
- && left.module.version === right.module.version
- && left.name === right.name;
-}
-
-function attributes(element: StructuredElement, required: readonly string[], optional: readonly string[] = []): void {
- const allowed = new Set([...required, ...optional]);
- const unknown = Object.keys(element.attributes).filter((name) => !allowed.has(name));
- if (unknown.length > 0 || required.some((name) => element.attributes[name] === undefined)) {
- throw new Error(
- `${element.name} requires ${required.join(", ")}`
- + (optional.length === 0 ? "" : `; optional: ${optional.join(", ")}`),
- );
- }
-}
-
-function stringAttribute(element: StructuredElement, name: string): string {
- const value = element.attributes[name];
- if (typeof value !== "string" || value.trim().length === 0) {
- throw new Error(`${element.name}.${name} must be a non-empty string`);
- }
- return value.trim();
-}
-
-function referencePath(element: StructuredElement, name: string): string {
- const value: MarkupAttributeValue | undefined = element.attributes[name];
- if (typeof value !== "object" || value.kind !== "reference" || value.path.length === 0) {
- throw new Error(`${element.name}.${name} must be a whole-value reference`);
- }
- return value.path;
-}
-
-function resolved(
- element: StructuredElement,
- name: string,
- resolveReference: (path: string) => SurfaceResolvedReference | undefined,
-): SurfaceResolvedReference {
- const path = referencePath(element, name);
- const value = resolveReference(path);
- if (value === undefined) throw new Error(`${element.name}.${name} cannot resolve ${path}`);
- return value;
-}
-
-function inline<T>(reference: SurfaceResolvedReference, subject: string): T {
- const value: StoredValue | undefined = reference.record?.value;
- if (value?.kind !== "inline") throw new Error(`${subject} must reference an authored inline value`);
- return value.value as unknown as T;
-}
-
-function mediaReference(reference: SurfaceResolvedReference, kind: "image" | "audio", subject: string) {
- if (!sameType(reference.type, artifactTypes.blob)) {
- throw new Error(`${subject} must reference a BlobArtifact`);
- }
- const value = reference.record?.value;
- if (value !== undefined && (value.kind !== "blob" || !value.mediaType.startsWith(`${kind}/`))) {
- throw new Error(`${subject} is not ${kind} media`);
- }
- return reference;
-}
-
-function recipeValue(
- element: StructuredElement,
- resolveReference: (path: string) => SurfaceResolvedReference | undefined,
-) {
- const reference = resolved(element, "recipe", resolveReference);
- if (!sameType(reference.type, svsRecipeType)) {
- throw new Error(`${element.name}.recipe must reference an SVS Recipe`);
- }
- const recipe = inline<SvsRecipe>(reference, `${element.name}.recipe`);
- if (recipe.contract !== "svml.svs-recipe@1") throw new Error(`${element.name}.recipe is invalid`);
- return recipe;
-}
-
-function kitValue(
- element: StructuredElement,
- resolveReference: (path: string) => SurfaceResolvedReference | undefined,
-): { readonly reference: SurfaceResolvedReference; readonly template: TextTemplate } {
- const reference = resolved(element, "kit", resolveReference);
- if (!sameType(reference.type, textTypes.template)) {
- throw new Error(`${element.name}.kit must reference a TextTemplate`);
- }
- const template = inline<TextTemplate>(reference, `${element.name}.kit`);
- verifyTextTemplate(template);
- return { reference, template };
-}
-
-function optionalTextValue(
- element: StructuredElement,
- name: string,
- resolveReference: (path: string) => SurfaceResolvedReference | undefined,
-): SurfaceResolvedReference | undefined {
- if (element.attributes[name] === undefined) return undefined;
- const reference = resolved(element, name, resolveReference);
- if (!sameType(reference.type, textTypes.text)) {
- throw new Error(`${element.name}.${name} must reference Text`);
- }
- const value = reference.record?.value;
- if (value !== undefined) {
- if (value.kind !== "inline") throw new Error(`${element.name}.${name} has an invalid authored Text value`);
- verifyText(value.value);
- }
- return reference;
-}
-
-function requiredTextValue(
- element: StructuredElement,
- name: string,
- resolveReference: (path: string) => SurfaceResolvedReference | undefined,
-) {
- const reference = resolved(element, name, resolveReference);
- if (!sameType(reference.type, textTypes.text)) {
- throw new Error(`${element.name}.${name} must reference Text`);
- }
- const value = reference.record?.value;
- if (value !== undefined) {
- if (value.kind !== "inline") throw new Error(`${element.name}.${name} has an invalid authored Text value`);
- verifyText(value.value);
- }
- return reference;
-}
-
-type ResolvedSpeakerReference = SpeakerReference & { readonly source: SurfaceResolvedReference };
-
-function references(
- element: StructuredElement,
- resolveReference: (path: string) => SurfaceResolvedReference | undefined,
-): ResolvedSpeakerReference[] {
- const result: ResolvedSpeakerReference[] = [];
- for (const child of element.children) {
- if (child.kind === "text") {
- if (child.value.trim().length > 0) throw new Error(`${element.name} accepts only Reference children`);
- continue;
- }
- if (localName(child.name) !== "Reference") throw new Error(`${element.name} accepts only Reference children`);
- attributes(child, [], ["image", "audio", "role"]);
- const kinds = (["image", "audio"] as const).filter((kind) => child.attributes[kind] !== undefined);
- if (kinds.length !== 1) throw new Error(`${child.name} requires exactly one of image or audio`);
- const kind = kinds[0]!;
- const source = mediaReference(resolved(child, kind, resolveReference), kind, `${child.name}.${kind}`);
- const rawRole = child.attributes.role;
- const role = rawRole === undefined
- ? kind === "image" ? "character-and-scene" : "voice-timbre"
- : typeof rawRole === "string" && rawRole.trim().length > 0
- ? rawRole.trim()
- : (() => { throw new Error(`${child.name}.role must be a non-empty string`); })();
- result.push({ kind, source, role });
- }
- return result;
-}
-
-function assembledReferences(
- id: string,
- endpoint: ExactModelEndpoint,
- values: readonly ResolvedSpeakerReference[],
- range: StructuredElement["range"],
-) {
- const mediaInputs: ExactModelMediaInput[] = [];
- const records: Array<{
- readonly id: string;
- readonly type: SurfaceResolvedReference["type"];
- readonly value: { readonly kind: "inline"; readonly value: CanonicalValue };
- readonly range: StructuredElement["range"];
- }> = [];
- const inputs: Record<string, SurfaceResolvedReference["ref"] | { readonly kind: "record"; readonly id: string }> = {};
- values.forEach((value, index) => {
- const name = `media-${String(index + 1).padStart(4, "0")}`;
- const portName = value.kind === "image" ? "referenceImage" : "referenceAudio";
- const endpointBinding = endpoint.mediaBindings[portName];
- if (endpointBinding === undefined) throw new Error(`${endpoint.ports.model} has no media port ${portName}`);
- const port = generationPort(endpoint.ports, portName);
- if (port.value.kind !== "media") throw new Error(`${endpoint.ports.model} port ${portName} is not media`);
- const bindingId = `${id}.${name}.binding`;
- records.push({
- id: bindingId,
- type: endpointBinding.type,
- value: {
- kind: "inline",
- value: sealGenerationMediaBinding(port as GenerationMediaPort, { role: value.kind }) as unknown as CanonicalValue,
- },
- range,
- });
- const names = exactModelMediaInputNames(name);
- inputs[names.binding] = { kind: "record", id: bindingId };
- inputs[names.artifact] = value.source.ref;
- mediaInputs.push({ name, port: portName });
- });
- return { mediaInputs, records, inputs };
-}
-
-function recipeString(
- recipe: SvsRecipe,
- name: string,
- fallback: string,
-): string {
- const value = recipe.properties[name] ?? fallback;
- if (typeof value !== "string" || value.trim().length === 0) {
- throw new Error(`Speaker Recipe ${recipe.path}.${name} must be a non-empty string`);
- }
- return value.trim();
-}
-
-function recipeBoolean(recipe: SvsRecipe, name: string, fallback: boolean): boolean {
- const value = recipe.properties[name] ?? fallback;
- if (typeof value !== "boolean") throw new Error(`Speaker Recipe ${recipe.path}.${name} must be boolean`);
- return value;
-}
-
-function model(recipe: SvsRecipe): SeedanceModel {
- const value = recipeString(recipe, "model", "mini");
- if (value === "mini" || value === "seedance-2-mini") return "seedance-2-mini";
- if (value === "fast" || value === "seedance-2-fast") return "seedance-2-fast";
- if (value === "standard" || value === "seedance-2") return "seedance-2";
- throw new Error(`Speaker Recipe ${recipe.path}.model is not supported`);
-}
-
-export const decodeSeedanceSpeakerTakeSurface: StructuredSurfaceHandler = ({ element, resolveReference }) => {
- attributes(element, ["id", "dialogue", "duration", "recipe", "kit"], ["action", "extra"]);
- const id = stringAttribute(element, "id");
- const kit = kitValue(element, resolveReference);
- const recipe = recipeValue(element, resolveReference);
- const reserved = new Set([
- "kind", "model", "resolution", "aspect-ratio", "web-search", "action", "extra",
- ]);
- if (recipeString(recipe, "kind", "ugc-talking-head") !== "ugc-talking-head") {
- throw new Error(`Speaker Recipe ${recipe.path}.kind must be ugc-talking-head`);
- }
- const selectedModel = model(recipe);
- const dialogue = requiredTextValue(element, "dialogue", resolveReference);
- const action = optionalTextValue(element, "action", resolveReference);
- const extra = optionalTextValue(element, "extra", resolveReference);
- const refs = references(element, resolveReference);
- for (const name of ["action", "extra"] as const) {
- if (recipe.properties[name] !== undefined) {
- throw new Error(`Speaker Recipe ${recipe.path}.${name} is content; pass it through ${element.name}.${name} as Text`);
- }
- }
- const promptParameters: Record<string, CanonicalValue> = {};
- for (const [name, value] of Object.entries(recipe.properties)) {
- if (reserved.has(name)) continue;
- if (typeof value !== "string" && typeof value !== "boolean" && !(typeof value === "number" && Number.isFinite(value))) {
- throw new Error(`Speaker Recipe ${recipe.path}.${name} must be a finite scalar`);
- }
- promptParameters[name] = value;
- }
- const intent = sealSpeakerTakeIntent({
- contract: "svml.seedance-speaker-take-intent@1",
- kit: kit.reference.path.split(".").at(-1) ?? "speaker-kit",
- recipe: { path: recipe.path },
- model: selectedModel,
- resolution: recipeString(recipe, "resolution", speakerMethodDefaults.resolution) as "480p" | "720p" | "1080p",
- aspectRatio: recipeString(recipe, "aspect-ratio", speakerMethodDefaults.aspectRatio) as "1:1" | "4:3" | "3:4" | "16:9" | "9:16" | "21:9" | "adaptive",
- webSearch: recipeBoolean(recipe, "web-search", speakerMethodDefaults.webSearch),
- promptParameters,
- references: refs.map(({ source: _source, ...reference }) => reference),
- });
- const bindings = createSpeakerTextBindings(intent);
- // Static template semantics remain an ordinary deterministic graph step.
- const program = createSpeakerDurationProgram(intent);
- const duration = resolved(element, "duration", resolveReference);
- if (!sameType(duration.type, speechTypes.duration)) {
- throw new Error(`${element.name}.duration must reference a SpeechDuration`);
- }
- const bindingsId = `${id}.bindings`;
- const programId = `${id}.program`;
- const endpoint = seedanceEndpointsByModel[selectedModel];
- const assembled = assembledReferences(id, endpoint, refs, element.range);
- const dynamicText = [
- { name: "dialogue", source: dialogue.ref },
- ...(action === undefined ? [] : [{ name: "action", source: action.ref }]),
- ...(extra === undefined ? [] : [{ name: "extra", source: extra.ref }]),
- ];
- const promptFragment = createTextRenderFragment(dynamicText.map(({ name }) => ({ name })));
- const generationFragment = createSeedanceSpeakerTakeFragment(
- endpoint,
- seedanceDurationCompileProducers[selectedModel],
- assembled.mediaInputs,
- );
- return {
- records: [{
- id: bindingsId,
- type: textTypes.bindings,
- value: { kind: "inline", value: bindings as unknown as CanonicalValue },
- range: element.range,
- }, {
- id: programId,
- type: seedanceTypes.durationProgram,
- value: { kind: "inline", value: program as unknown as CanonicalValue },
- range: element.range,
- }, ...dynamicText.map(({ name }) => ({
- id: `${id}.${name}-binding`,
- type: textTypes.binding,
- value: {
- kind: "inline" as const,
- value: sealTextBinding({ name, mode: "set" }) as unknown as CanonicalValue,
- },
- range: element.range,
- })), ...assembled.records],
- components: [{
- id: `${id}.render-prompt`,
- fragment: promptFragment.id,
- inputs: {
- template: kit.reference.ref,
- bindings: { kind: "record", id: bindingsId },
- ...Object.fromEntries(dynamicText.flatMap(({ name, source }) => [[
- `binding:${name}:spec`, { kind: "record" as const, id: `${id}.${name}-binding` },
- ], [
- `binding:${name}:text`, source,
- ]])),
- },
- outputs: { text: `${id}.prompt` },
- range: element.range,
- }, {
- id,
- fragment: generationFragment.id,
- inputs: {
- program: { kind: "record", id: programId },
- duration: duration.ref,
- [exactModelTextInputName("prompt")]: {
- kind: "component-output",
- component: `${id}.render-prompt`,
- output: "text",
- },
- ...assembled.inputs,
- },
- outputs: {
- video: `${id}.video`,
- },
- range: element.range,
- }],
- fragments: [promptFragment, generationFragment],
- };
-};
diff --git a/packages/seedance-speaker/src/types.ts b/packages/seedance-speaker/src/types.ts
deleted file mode 100644
index 8a436e8c..00000000
--- a/packages/seedance-speaker/src/types.ts
+++ /dev/null
@@ -1,19 +0,0 @@
-import type { CanonicalValue } from "@narratage/protocol";
-import type { SeedanceModel } from "@narratage/seedance";
-
-export type SpeakerReference = {
- readonly kind: "image" | "audio";
- readonly role: string;
-};
-
-export type SpeakerTakeIntent = {
- readonly contract: "svml.seedance-speaker-take-intent@1";
- readonly kit: string;
- readonly recipe: { readonly path: string };
- readonly model: SeedanceModel;
- readonly resolution: "480p" | "720p" | "1080p";
- readonly aspectRatio: "1:1" | "4:3" | "3:4" | "16:9" | "9:16" | "21:9" | "adaptive";
- readonly webSearch: boolean;
- readonly promptParameters: Readonly<Record<string, CanonicalValue>>;
- readonly references: readonly SpeakerReference[];
-};
diff --git a/packages/seedance-speaker/test/kit.test.ts b/packages/seedance-speaker/test/kit.test.ts
deleted file mode 100644
index 839ac82d..00000000
--- a/packages/seedance-speaker/test/kit.test.ts
+++ /dev/null
@@ -1,86 +0,0 @@
-import assert from "node:assert/strict";
-import { readFileSync } from "node:fs";
-import test from "node:test";
-
-import {
- bindText,
- renderText,
- sealText,
- sealTextBinding,
- textTemplateFromSvsRecipes,
- verifyText,
-} from "@narratage/text";
-import { parseSvs } from "@narratage/svs";
-import {
- createSpeakerDurationProgram,
- createSpeakerTextBindings,
- sealSpeakerTakeIntent,
- speakerMethodDefaults,
-} from "@narratage/seedance-speaker";
-
-const officialUgcV1SvsSource = readFileSync(new URL("../kits/official-ugc-v1.svs", import.meta.url), "utf8");
-const officialUgcV1Parsed = parseSvs(
- "official-ugc-v1.svs",
- officialUgcV1SvsSource.slice(officialUgcV1SvsSource.indexOf("<sheet")),
-);
-const officialUgcV1Template = textTemplateFromSvsRecipes(
- officialUgcV1Parsed.recipes.map((item) => item.value),
- "official-ugc-v1",
-);
-
-function intent() {
- return sealSpeakerTakeIntent({
- contract: "svml.seedance-speaker-take-intent@1",
- kit: "official-ugc-v1",
- recipe: { path: "speaker.default" },
- model: speakerMethodDefaults.model,
- resolution: speakerMethodDefaults.resolution,
- aspectRatio: speakerMethodDefaults.aspectRatio,
- webSearch: speakerMethodDefaults.webSearch,
- promptParameters: {},
- references: [
- { kind: "image", role: "character-and-scene" },
- { kind: "audio", role: "voice-timbre" },
- ],
- });
-}
-
-test("official UGC Kit renders one ordinary ordered Text value", () => {
- const bindings = bindText(
- createSpeakerTextBindings(intent()),
- sealTextBinding({ name: "dialogue", mode: "set" }),
- sealText("HOST: Say exactly these words."),
- );
- const prompt = renderText(officialUgcV1Template, bindings);
- verifyText(prompt);
- assert.match(prompt.value, /^Create a realistic vertical iPhone UGC/u);
- assert.match(prompt.value, /@audio1/u);
- assert.match(prompt.value, /SCRIPT TO DELIVER:\nHOST: Say exactly these words\./u);
- assert.ok(prompt.value.indexOf("COMPOSITION STABILITY") < prompt.value.indexOf("CAMERA MOTION"));
-});
-
-test("Speaker method preserves exact Seedance settings while prompt and media stay on graph edges", () => {
- const take = intent();
- const program = createSpeakerDurationProgram(take);
- assert.equal(program.model, "seedance-2-mini");
- assert.deepEqual(program.ports.resolution, ["720p"]);
- assert.deepEqual(program.ports.aspectRatio, ["9:16"]);
- assert.equal(program.ports.referenceImage, undefined);
- assert.equal(program.ports.referenceAudio, undefined);
- assert.equal(program.ports.prompt, undefined, "prompt arrives from an explicit Text edge");
- assert.equal(program.ports.duration, undefined, "duration arrives from speech estimation");
-});
-
-test("Speaker rejects invalid media while Text Template rejects unknown parameters", () => {
- const base = intent();
- assert.throws(() => sealSpeakerTakeIntent({
- ...base,
- references: [{ kind: "audio", role: "voice-timbre" }],
- }), /requires 1-9 image references/u);
- assert.throws(() => sealSpeakerTakeIntent({ ...base, resolution: "1080p" }), /at most 720p/u);
- const unknown = sealSpeakerTakeIntent({ ...base, promptParameters: { "imaginary-axis": "value" } });
- assert.throws(
- () => renderText(officialUgcV1Template, createSpeakerTextBindings(unknown)),
- /binding imaginary-axis is not declared/u,
- );
-});
diff --git a/packages/video-cli/test/cli.test.ts b/packages/video-cli/test/cli.test.ts
index da3a8892..d8c0ccf2 100644
--- a/packages/video-cli/test/cli.test.ts
+++ b/packages/video-cli/test/cli.test.ts
@@ -187,37 +187,49 @@ test("an installed open-font package supplies a multilingual exact stack without
);
});
-test("Text Template and Speaker keep prompt assembly visible in the Run Graph", async () => {
+test("Text Template and exact Seedance keep speaker prompt assembly visible in the Run Graph", async () => {
const root = await mkdtemp(join(tmpdir(), "svml-cli-text-template-"));
const file = join(root, "main.svml");
const kitSource = await readFile(
- new URL("../../seedance-speaker/kits/official-ugc-v1.svs", import.meta.url),
+ new URL("../../seedance-kits/kits/speaker-v1.svs", import.meta.url),
"utf8",
);
- await writeFile(join(root, "official-ugc-v1.svs"), kitSource, "utf8");
+ await writeFile(join(root, "speaker-v1.svs"), kitSource, "utf8");
await writeFile(join(root, "studio.svs"), `<?svml using="@narratage/svs@1"?>
<sheet version="1">
speech.normal { language: en; pace: normal; min: 4; max: 15; rounding: round; }
- speaker.default { kind: ugc-talking-head; model: mini; resolution: 720p; aspect-ratio: 9:16; }
+ speaker.default {
+ composition-stability: soft-locked; camera-motion: none;
+ edit-rhythm: continuous-take; performance: natural-explainer; gesture: natural;
+ }
</sheet>`, "utf8");
await writeFile(join(root, "host.png"), new Uint8Array([137, 80, 78, 71]));
await writeFile(file, `<?svml using="@narratage/markup@1"?>
<svml>
<import from="@narratage/script@1"/>
+ <import as="text" from="@narratage/text@1"/>
<import as="media" from="@narratage/media@1"/>
<import as="estimate" from="@narratage/estimate@1"/>
- <import as="speaker" from="@narratage/seedance-speaker@1"/>
+ <import as="seedance" from="@narratage/seedance@1"/>
<import as="studio" source="./studio.svs"/>
- <import as="ugc" source="./official-ugc-v1.svs"/>
+ <import as="speaker-kit" source="./speaker-v1.svs"/>
<script id="story"><opening><HOST>Meaning becomes the source.</opening></script>
<media:Image id="host" src="./host.png"/>
+ <media:Audio id="voice" src="./voice.mp3"/>
+ <text:Value id="action">Hold direct eye contact and use one compact gesture.</text:Value>
<estimate:Speech id="duration" source={story.segment.opening.speech} policy={studio.speech.normal}/>
- <speaker:Take id="take" dialogue={story.segment.opening.dialogue}
- duration={duration.duration} recipe={studio.speaker.default} kit={ugc.official-ugc-v1}>
- <speaker:Reference image={host} role="character-and-scene"/>
- </speaker:Take>
+ <text:Render id="take-prompt" template={speaker-kit.speaker-v1} recipe={studio.speaker.default}>
+ <text:Set name="dialogue" text={story.segment.opening.dialogue}/>
+ <text:Set name="action" text={action}/>
+ </text:Render>
+ <seedance:ReferenceVideo id="take" model="mini" prompt={take-prompt}
+ duration={duration.duration} resolution="720p" aspect-ratio="9:16" generate-audio="true">
+ <seedance:Reference image={host}/>
+ <seedance:Reference audio={voice}/>
+ </seedance:ReferenceVideo>
</svml>`, "utf8");
+ await writeFile(join(root, "voice.mp3"), new Uint8Array([73, 68, 51]));
let checkedOutput = "";
await runCli(["check", file], { write: (text) => { checkedOutput += text; } });
@@ -225,8 +237,7 @@ test("Text Template and Speaker keep prompt assembly visible in the Run Graph",
readonly exports: readonly { readonly name: string }[];
};
const names = checked.exports.map((item) => item.name);
- assert.equal(names.includes("take.prompt"), true);
- assert.equal(names.includes("take.program"), true);
+ assert.equal(names.includes("take-prompt"), true);
assert.equal(names.includes("take.video"), true);
const run = await writeRun(root, "build.svrun", [{ output: "take.video" }]);
@@ -237,8 +248,10 @@ test("Text Template and Speaker keep prompt assembly visible in the Run Graph",
};
assert.deepEqual(plan.steps.map((step) => step.producer.name).sort(), [
"bind-request-seedance-2-mini-referenceImage",
+ "bind-request-seedance-2-mini-referenceAudio",
"bind-request-seedance-2-mini-prompt-text",
"bind-text",
+ "bind-text",
"compile-seedance-2-mini-duration-request",
"estimate-speech-duration",
"finalize-request-seedance-2-mini",
diff --git a/packages/video-cli/test/packages.ts b/packages/video-cli/test/packages.ts
index a72c5465..43fbedd8 100644
--- a/packages/video-cli/test/packages.ts
+++ b/packages/video-cli/test/packages.ts
@@ -17,7 +17,6 @@ import programSpace from "../../program-space/src/activation.js";
import text from "../../text/src/activation.js";
import runMarkup from "../../run-markup/src/activation.js";
import script from "../../script/src/activation.js";
-import seedanceSpeaker from "../../seedance-speaker/src/activation.js";
import seedance from "../../seedance/src/activation.js";
import speechAlignment from "../../speech-alignment/src/activation.js";
import speechSpine from "../../speech-spine/src/activation.js";
@@ -50,7 +49,6 @@ export const videoTestPackages = [
text,
generation,
seedance,
- seedanceSpeaker,
caption,
captionFine,
captionGemini,
diff --git a/packages/whisperx/README.md b/packages/whisperx/README.md
index 072d3a15..9af5b809 100644
--- a/packages/whisperx/README.md
+++ b/packages/whisperx/README.md
@@ -2,7 +2,7 @@
Explicit WhisperX model-family capability for the official speech program. Importing this package
selects WhisperX; Runtime registration only binds the resulting `WhisperXAlignmentEvidence` Need to
-a concrete WhisperX execution endpoint such as a trusted local worker or Hypit.
+a concrete WhisperX execution endpoint such as a trusted local worker or hosted service.
The package contains no credentials, Python environment or queue. Providers translate the typed
request constraints and return the canonical WhisperX result contract. A deterministic Producer
diff --git a/pnpm-lock.yaml b/pnpm-lock.yaml
index 72c6b44d..e0ec9c66 100644
--- a/pnpm-lock.yaml
+++ b/pnpm-lock.yaml
@@ -128,9 +128,6 @@ importers:
'@narratage/seedance-kits':
specifier: workspace:*
version: link:packages/seedance-kits
- '@narratage/seedance-speaker':
- specifier: workspace:*
- version: link:packages/seedance-speaker
'@narratage/semantic-map':
specifier: workspace:*
version: link:packages/semantic-map
@@ -2157,39 +2154,6 @@ importers:
packages/seedance-kits: {}
- packages/seedance-speaker:
- dependencies:
- '@narratage/artifact':
- specifier: workspace:*
- version: link:../artifact
- '@narratage/elaborator':
- specifier: workspace:*
- version: link:../elaborator
- '@narratage/generation':
- specifier: workspace:*
- version: link:../generation
- '@narratage/markup':
- specifier: workspace:*
- version: link:../markup
- '@narratage/model-kit':
- specifier: workspace:*
- version: link:../model-kit
- '@narratage/protocol':
- specifier: workspace:*
- version: link:../protocol
- '@narratage/seedance':
- specifier: workspace:*
- version: link:../seedance
- '@narratage/speech':
- specifier: workspace:*
- version: link:../speech
- '@narratage/svs':
- specifier: workspace:*
- version: link:../svs
- '@narratage/text':
- specifier: workspace:*
- version: link:../text
-
packages/seedream:
dependencies:
'@narratage/generation':
diff --git a/spec/audio-track.md b/spec/audio-track.md
index 514a31d9..d9e00658 100644
--- a/spec/audio-track.md
+++ b/spec/audio-track.md
@@ -208,7 +208,7 @@ not a privileged lane. A package that naturally owns one visual hit and one soun
`VisualTrack` and `AudioTrack` outputs from one Fragment, while Film receives both through ordinary
edges.
-## 8. Twinit migration audit
+## 8. Legacy migration audit
Retain:
diff --git a/spec/caption-fine.md b/spec/caption-fine.md
index 64e99e10..37cfcb7e 100644
--- a/spec/caption-fine.md
+++ b/spec/caption-fine.md
@@ -139,7 +139,7 @@ Activation state and decoration geometry are deliberately separate. `current` ac
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 old Twinit behavior—trail text with a current-only pill—is directly expressible.
+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
diff --git a/spec/deck-track.md b/spec/deck-track.md
index 79c9eea1..d0f968fd 100644
--- a/spec/deck-track.md
+++ b/spec/deck-track.md
@@ -229,7 +229,7 @@ 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 Twinit
+## 10. Migration from the legacy implementation
Retain from the old Deck implementation:
@@ -265,7 +265,7 @@ Retire:
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 Twinit-style
+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
diff --git a/spec/media-track.md b/spec/media-track.md
index b40c9b80..48085769 100644
--- a/spec/media-track.md
+++ b/spec/media-track.md
@@ -532,7 +532,7 @@ 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.
-## 14. Twinit audit: retained and retired
+## 14. Legacy audit: retained and retired
Retain:
diff --git a/spec/ranking-track.md b/spec/ranking-track.md
index 02f87b9c..13c7241c 100644
--- a/spec/ranking-track.md
+++ b/spec/ranking-track.md
@@ -3,7 +3,7 @@
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 Twinit node format or a frozen public ABI.
+historical node format or a frozen public ABI.
## 1. What Ranking means
diff --git a/spec/screen-overlay.md b/spec/screen-overlay.md
index 1fbc3915..42eba046 100644
--- a/spec/screen-overlay.md
+++ b/spec/screen-overlay.md
@@ -199,7 +199,7 @@ 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. Twinit migration audit
+## 8. Legacy migration audit
Retain:
diff --git a/spec/script-surface.md b/spec/script-surface.md
index 261b0316..e4e1074c 100644
--- a/spec/script-surface.md
+++ b/spec/script-surface.md
@@ -62,21 +62,21 @@ Script 编译必须先生成保留源码映射的 Narrative IR,再从同一份
</script>
```
-若 `product = "Hypit"`、`product_pronunciation = "high pit"`,三个文本投影
+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 high pit.
+B: Meet ess vee em ell.
speech
I just laughed my ass out.
-Meet high pit.
+Meet ess vee em ell.
caption
I just lmao.
-Meet Hypit.
+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
diff --git a/spec/spatial-layout.md b/spec/spatial-layout.md
index b0465f4a..072993ab 100644
--- a/spec/spatial-layout.md
+++ b/spec/spatial-layout.md
@@ -408,7 +408,7 @@ does not gain source, role, timing or Artifact digests. `VisualTrack` does not g
or Text fields. Affinity is validated by the multi-input fitting/lowering Operation rather than by
copying lineage metadata through every value.
-## 12. Twinit findings retained and retired
+## 12. Legacy findings retained and retired
Retained:
diff --git a/spec/typography-track.md b/spec/typography-track.md
index 03b859d8..216b3247 100644
--- a/spec/typography-track.md
+++ b/spec/typography-track.md
@@ -58,9 +58,9 @@ It currently cannot represent:
This is a useful end-to-end witness only. Extending `TextAppearance` field by field would preserve
the wrong model.
-### 2.2 Twinit evidence
+### 2.2 Legacy evidence
-Twinit's later `TextLayoutV2` discovered several correct separations:
+The legacy `TextLayoutV2` discovered several correct separations:
- inline/block `hug | fixed` sizing;
- independent padding and inline/block alignment;
diff --git a/tools/graph-first-value-boundary.test.mjs b/tools/graph-first-value-boundary.test.mjs
index eb540cde..e371d083 100644
--- a/tools/graph-first-value-boundary.test.mjs
+++ b/tools/graph-first-value-boundary.test.mjs
@@ -42,7 +42,6 @@ const domainValuePackages = [
"text",
"script",
"seedance",
- "seedance-speaker",
"speech-alignment",
"speech-spine",
"speech-basis",
diff --git a/tools/package-boundaries.test.mjs b/tools/package-boundaries.test.mjs
index 555b3da8..077fbcf9 100644
--- a/tools/package-boundaries.test.mjs
+++ b/tools/package-boundaries.test.mjs
@@ -185,7 +185,6 @@ test("generic and video CLIs reach no Provider package and video CLI activates n
const videoDependencies = transitive(graph, "@narratage/video-cli");
assert.deepEqual(providers.filter((name) => videoDependencies.has(name)), []);
assert.ok(!videoDependencies.has("@narratage/script"));
- assert.ok(!videoDependencies.has("@narratage/seedance-speaker"));
assert.ok(!videoDependencies.has("@narratage/media-track"));
assert.ok(!videoDependencies.has("@narratage/typography-track"));
assert.ok(!videoDependencies.has("@narratage/film"));
@@ -196,7 +195,7 @@ test("domain packages confine their Markup dependency to Surface and activation
const { readdir, readFile } = await import("node:fs/promises");
const surfaceOnly = [
"caption", "caption-gemini", "estimate", "film", "render-hyperframes",
- "image-transform", "media", "media-pipeline", "media-track", "seedance", "seedance-speaker", "speech-spine", "whisperx",
+ "image-transform", "media", "media-pipeline", "media-track", "seedance", "speech-spine", "whisperx",
];
const allowed = new Set(["surface.ts", "activation.ts"]);
for (const name of surfaceOnly) {
diff --git a/tools/repository-hygiene.test.mjs b/tools/repository-hygiene.test.mjs
index ac55a2d5..0659ac53 100644
--- a/tools/repository-hygiene.test.mjs
+++ b/tools/repository-hygiene.test.mjs
@@ -6,12 +6,25 @@ import ts from "typescript";
const repositoryRoot = new URL("../", import.meta.url);
const scanRoots = ["docs", "examples", "packages", "services", "spec", "test", "tools"];
+const rootTextFiles = ["README.md", "package.json", "pnpm-workspace.yaml", "tsconfig.json"];
const ignoredDirectories = new Set(["node_modules", "dist", "output", ".svml", ".vitepress"]);
const textExtensions = new Set([
".css", ".html", ".js", ".json", ".jsx", ".md", ".mjs", ".sh", ".svml", ".svrun",
".svs", ".ts", ".tsx", ".txt", ".yaml", ".yml",
]);
+function isMountedDocumentationPath(path) {
+ return path === "docs/public"
+ || path.startsWith("docs/public/")
+ || path === "docs/index.md"
+ || path === "docs/quickstart.md"
+ || path.startsWith("docs/quickstart/")
+ || path === "docs/guide"
+ || path.startsWith("docs/guide/")
+ || path === "docs/zh"
+ || path.startsWith("docs/zh/");
+}
+
async function repositoryEntries() {
const entries = [];
const visit = async (url) => {
@@ -27,17 +40,27 @@ async function repositoryEntries() {
return entries;
}
+async function repositoryTextEntries() {
+ const entries = (await repositoryEntries()).filter((entry) =>
+ entry.isFile && textExtensions.has(extname(entry.path)));
+ for (const path of rootTextFiles) {
+ entries.push({ path, child: new URL(`../${path}`, import.meta.url), isFile: true });
+ }
+ return entries;
+}
+
function lineOf(source, offset) {
return source.slice(0, offset).split("\n").length;
}
-test("public repository contains no customer or product brand residue", async () => {
+test("source distribution outside the mounted documentation contains no customer or product brand residue", async () => {
const brand = new RegExp(
- String.raw`\b\x65\x63\x68\x6f(?:[\s_-]+)\x70\x72\x6f\b|\b\x65\x63\x68\x6f\x70\x72\x6f\b|\bwith[-_]\x65\x63\x68\x6f\b`,
+ String.raw`\b\x65\x63\x68\x6f(?:[\s_-]+)\x70\x72\x6f\b|\b\x65\x63\x68\x6f\x70\x72\x6f\b|\bwith[-_]\x65\x63\x68\x6f\b|\b\x61\x72\x65\x75\x6d\b|\b\x6c\x6f\x6f\x6b\x73\x6d\x61\x78\b`,
"giu",
);
const failures = [];
for (const entry of await repositoryEntries()) {
+ if (isMountedDocumentationPath(entry.path)) continue;
if (brand.test(entry.path)) failures.push(`${entry.path} (path)`);
brand.lastIndex = 0;
if (!entry.isFile || !textExtensions.has(extname(entry.path))) continue;
@@ -47,6 +70,36 @@ test("public repository contains no customer or product brand residue", async ()
assert.deepEqual(failures, [], `brand residue found:\n${failures.join("\n")}`);
});
+test("source distribution outside the mounted documentation contains no retired project identities", async () => {
+ const retired = new RegExp(
+ String.raw`\b\x74\x77\x69\x6e\x69\x74\b|\b\x68\x79\x70\x69\x74\b|\b\x6d\x65\x67\x6e\x65\x74\x61\b`,
+ "giu",
+ );
+ const failures = [];
+ for (const entry of await repositoryTextEntries()) {
+ if (isMountedDocumentationPath(entry.path)) continue;
+ const source = await readFile(entry.child, "utf8");
+ for (const match of source.matchAll(retired)) failures.push(`${entry.path}:${lineOf(source, match.index)}`);
+ }
+ assert.deepEqual(failures, [], `retired internal identity found:\n${failures.join("\n")}`);
+});
+
+test("public repository contains no workstation paths or high-confidence secret bytes", async () => {
+ const workstationPath = /(?:\/Users\/[A-Za-z0-9._-]+\/|\/home\/[A-Za-z0-9._-]+\/|[A-Za-z]:\\Users\\[^\\]+\\)/gu;
+ const secretBytes = /(?:AKIA[0-9A-Z]{16}|AIza[0-9A-Za-z_-]{30,}|-----BEGIN (?:RSA |EC |OPENSSH )?PRIVATE KEY-----)/gu;
+ const failures = [];
+ for (const entry of await repositoryTextEntries()) {
+ const source = await readFile(entry.child, "utf8");
+ for (const match of source.matchAll(workstationPath)) {
+ failures.push(`${entry.path}:${lineOf(source, match.index)} (workstation path)`);
+ }
+ for (const match of source.matchAll(secretBytes)) {
+ failures.push(`${entry.path}:${lineOf(source, match.index)} (secret bytes)`);
+ }
+ }
+ assert.deepEqual(failures, [], `private repository residue found:\n${failures.join("\n")}`);
+});
+
test("pre-release project-owned major identities remain at one", async () => {
const identity = /(?:\bsvml\.[a-z0-9._/-]+|@narratage\/[a-z0-9._/${}-]+)@([0-9]+)(?![.0-9])/giu;
const failures = [];