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.
TypeScript agentic-aiaiai-agentsai-videocompilerdslffmpeggenerative-aillmmarkup-languagemonorepoplugin-systemprogramming-languagetext-to-videotypescriptvideovideo-automationvideo-clonevideo-editingvideo-generation
docs(skill): align performance and visual evidence guidance
kashorind3aec68fcf10
diff --git a/packages/image-compose/README.md b/packages/image-compose/README.mdindex 4c53953c..2922c8e8 100644--- a/packages/image-compose/README.md+++ b/packages/image-compose/README.md@@ -1,14 +1,25 @@# @hypit/image-compose-Deterministic multi-layer raster composition for SVML.+Deterministically flatten a fixed arrangement of existing still images into one PNG.The author graph supplies an explicit Canvas, an ordered list of image Layers, and an explicit Frame for every Layer. The package lowers that meaning to the shared `@hypit/raster` execution contract; an Endpoint such as `@hypit/provider-image-opencv-local` performs the pixels. Core knows nothing about images or compositing.```xml-<compose:Image id="card" canvas={portrait} background="#00000000">- <compose:Layer source={background.image} frame={full} fit="cover"/>- <compose:Layer source={product.image} frame={productFrame} fit="contain"/>+<import as="space" from="@hypit/spatial@1"/>+<import as="compose" from="@hypit/image-compose@1"/>++<space:Canvas id="comparison-canvas" width="2048" height="1024"/>+<space:Frame id="before-panel" within={comparison-canvas}+ left="0%" top="0%" right="50%" bottom="100%"/>+<space:Frame id="after-panel" within={comparison-canvas}+ left="50%" top="0%" right="100%" bottom="100%"/>++<compose:Image id="comparison" canvas={comparison-canvas} background="#EEEAE2FF">+ <compose:Layer source={before.image} frame={before-panel} fit="contain"/>+ <compose:Layer source={after.image} frame={after-panel} fit="contain"/></compose:Image>```Child order is paint order. Frames may extend beyond the Canvas and are clipped. Version 1 always produces a PNG using normal alpha compositing; it deliberately has no implicit base image, layout, or metadata propagation.++This operation does no visual reasoning. It does not reconcile perspective, depth, light, subject scale, scene continuity or a natural seam between its inputs. Use an image-generation or image-editing model when the desired result should become one newly directed camera image. Keep independently timed or editable layers in Tracks and Film. Image Compose is for the narrower case where the intended result is precisely the fixed two-dimensional arrangement itself.diff --git a/packages/image-compose/src/manifest.ts b/packages/image-compose/src/manifest.tsindex 864482f2..19a53892 100644--- a/packages/image-compose/src/manifest.ts+++ b/packages/image-compose/src/manifest.ts@@ -48,7 +48,7 @@ export const imageComposeMarkupSurfaces = [{name: "image", tag: "Image", mode: "structured",outputs: [imageComposeTypes.options, imageComposeTypes.layerSpec, artifactTypes.blob],vocabulary: {- summary: "Paints ordered image Layers onto one Canvas and publishes the composed picture as an image Artifact.",+ summary: "Flattens ordered still-image Layers at explicit rectangular Frames into one PNG Artifact.",attributes: [{ name: "id", kind: "identifier", required: true,summary: "Names this composition so its image can be referenced elsewhere in the Source." },@@ -78,14 +78,15 @@ export const imageComposeMarkupSurfaces = [{{ name: "image", type: artifactTypes.blob,summary: "The composed picture, a PNG." },],- example: `<compose:Image id="card" canvas={portrait} background="#00000000">- <compose:Layer source={background.image} frame={full} fit="cover"/>- <compose:Layer source={product.image} frame={product-frame} fit="contain"/>+ example: `<compose:Image id="comparison" canvas={comparison-canvas} background="#EEEAE2FF">+ <compose:Layer source={before.image} frame={before-panel} fit="contain"/>+ <compose:Layer source={after.image} frame={after-panel} fit="contain"/></compose:Image>`,notes: ["`background` is written as `#RRGGBBAA` and defaults to `#00000000`.","The composition requires at least one Layer and holds at most 64.","Child order is paint order, and a Frame that extends beyond the Canvas is clipped.",+ "The operation performs rectangular raster placement only; it does not redesign a scene or create timed layers.",],},}] as const;diff --git a/skills/hypit/SKILL.md b/skills/hypit/SKILL.mdindex 56ba3179..1ab58e49 100644--- a/skills/hypit/SKILL.md+++ b/skills/hypit/SKILL.md@@ -38,31 +38,22 @@ crafts explain resolution and duration choices.## Establish the working crew-Before promising a generated reconstruction, establish both sides of the work: the ability to-understand the reference and the ability to produce the new piece. Inspect the selected Runtime-Profile and its Endpoint diagnoses first. A typical reconstruction needs Gemini for visual-observation, WhisperX for word transcription and alignment, an image-generation Endpoint—usually-GPT Image 2—and a video-generation Endpoint—usually Seedance 2 Mini. The Treatment may demand-additional voice or audio capabilities.--Continue without interrupting the user when those capabilities are already reachable. When they are-not, explain all material gaps together and ask which supported Provider accounts or Keys the user-already has and whether preparing local WhisperX is practical. Keep secret values in `hypit auth`,-not in conversation. Offer HypiHub as the official hosted choice when the user wants one account for-observation, alignment, image generation, and video generation, or does not have the corresponding-BYOK or local capabilities. This is an explicit environment choice; the Runtime Profile reference-owns setup and verification.--State the actual consequence of anything still missing. Without a working eye and ear, do not claim-to have faithfully reconstructed the reference. Without the required image or video Endpoint, do not-promise generated A-roll or B-roll. HyperFrames can still generate a deterministic MG-, Caption-, or-Typography-led video when that is the work the user wants.+At the beginning of reference reconstruction, read the+[Runtime Profile](references/environment/profile.md) and establish the capabilities needed to+understand the supplied work and produce the new piece: moving-image observation, word alignment for+spoken work, and the image, video, voice or audio generation demanded by the Treatment. Continue when+they are reachable. When they are not, present the material gaps together, explain their consequence+for this work, and let the user choose among the real account, local and hosted options before relying+on the missing capability.## Two loops, one directorUnderstanding a reference is a loop between coarse and fine: the whole-piece reading tells your eyesand ears where a closer look would change the work, and close evidence rewrites the whole-piece-reading.+reading. A limitation accepted earlier does not freeze the work at that level. When a newly reachable+capability or a new Result can materially improve an earlier judgment, return to that judgment and+upgrade it before relying on it for further creative or paid work. Preserve what remains sound and+revise what the new evidence changes.Making the new piece is a loop between intention and result: the Brief holds what the user asked for,the Treatment is your directorial answer, and Source, Recipe, Run, and components make it exact.@@ -74,16 +65,22 @@ design itself changes. Brief changes when the user's goal or constraints change.Create project components as normal production work. Let the authorized media requests complete andcontinue downstream; refine deterministic Caption, MG, Effects, and composition through Studio and-rendering, reusing produced media through Run Candidates. Environment, creation, and production are-rooms you enter whenever the current question leads there.+rendering, reusing produced media through Run Candidates. Supplied and produced images and videos+are the primary visual evidence for the work. A preview substitute is a short-lived answer to a+current composition or wiring question, not a required stage, approval gate, or production state;+select the most representative available media as soon as it exists. Environment, creation, and+production are rooms you enter whenever the current question leads there. They are not stages that+close behind the work.## Standing responsibilities- **Money.** Before paid work, including observation and transcription, explain the actual requests,- selected Endpoints, and Provider price sources; state plainly when a price cannot be derived. For a Build,- use `hypit plan` with the selected Runtime Profile. Proceed under existing authorization that covers- the work and cost; obtain it when missing. Additional generation, including a repeated attempt,- needs an explicit decision covered by that authority. A request to rebuild or fix downstream work+ selected Endpoints, and the current Provider pricing information available for them. Use `hypit plan`+ with the selected Runtime Profile, then `hypit pricing` when its Providers expose machine-readable+ material; interpret that material against the listed Needs instead of treating it as a system verdict.+ Pricing availability is not a Build gate. Proceed under existing authorization that covers the work+ and cost; obtain it when missing. Additional generation needs an explicit decision covered by that+ authority. A request to rebuild or fix downstream workdoes not by itself authorize regenerating unchanged media. A Provider or billing-account change isa separate user choice; the Runtime Profile reference below owns how that choice is made.- **Existing work.** When revising or retrying, preserve usable produced media through explicit Run@@ -138,7 +135,7 @@ rooms you enter whenever the current question leads there.| Run syntax, Targets, Candidates, Run Fragments or substitute media for a preview | `references/production/runs.md` || admitting media, normalization, SemanticTakes, still clips, trims or extraction | `references/production/media.md` || composing, correcting, resizing, cropping or cutting out an image | `references/production/image-operations.md` |-| insets, layered screens, or a subject whose background needs removing | `references/playbooks/craft/compositing.md` |+| faces, MG, Caption, insets or layered screens must share one frame, or a subject's background needs removing | `references/playbooks/craft/compositing.md` || Canvas, Frames, aspect, fitting, cropping or coordinate relationships | `references/production/spatial.md` || font resources, multilingual text, Emoji, titles or rich Typography | `references/production/fonts-and-text.md` || which installed Surface to use, or whether to write a project component | `references/production/vocabulary.md` |@@ -147,7 +144,7 @@ rooms you enter whenever the current question leads there.| writing a project Track with new layout, semantic events or persistent state | `references/production/track-authoring.md` || drawing a component's elements, animation, resources or prepared surfaces | `references/production/component-visuals.md` || writing a Caption family with new word relationships, scheduling or layout | `references/production/caption-authoring.md` |-| `plan`, `build`, a retry or interrupted submission, following work, Results and exports | `references/production/builds.md` |+| `plan`, Provider pricing information, `build`, a retry or interrupted submission, following work, Results and exports | `references/production/builds.md` || opening Studio, understanding its views and edits, or adding a component Companion | `references/production/studio.md` || Film assembly, picture and sound, final rendering or a selected frame interval | `references/production/rendering.md` || judging the preview or finished Result and deciding what to fix | `references/production/review.md` |diff --git a/skills/hypit/references/creation/reference-video.md b/skills/hypit/references/creation/reference-video.mdindex 8f42019a..46f6a362 100644--- a/skills/hypit/references/creation/reference-video.md+++ b/skills/hypit/references/creation/reference-video.md@@ -5,15 +5,17 @@ from.## Direct the investigation-The Agent is the mind and director. Use visual observation as eyes, transcription and alignment as-ears, and local media commands to prepare a closer view. The tools report evidence; the Agent connects-events across time, resolves contradictions, explains why they matter, and decides what to inspect-next.--Begin by watching the whole piece. Spoken language carries most work Hypit is asked to clone, so obtain-and read its word-level transcript and use it as the normal time spine. If the piece is genuinely-speechless, establish that spine from its actions, visual changes, music, effects and silence. Form a-provisional account of:+The Agent is the mind and director. Moving-image observation is its broad eye, transcription and+alignment are its ears, and local media commands prepare closer views that the Agent can inspect+itself. The tools report evidence; the Agent connects events across time, resolves contradictions,+explains why they matter, and decides what to inspect next. Gemini supplies moving-image observations;+the Agent connects them into a whole-piece understanding and authors the target work.++Begin by watching the whole piece. For ordinary reconstruction, give the whole work or meaningful+spans to a capable moving-image observer so motion, order and distant relationships are actually in+evidence. Spoken language carries most work Hypit is asked to clone, so obtain and read its word-level+transcript and use it as the normal time spine. If the piece is genuinely speechless, establish that+spine from its actions, visual changes, music, effects and silence. Form a provisional account of:- the Hook, story movement, payoff, and intended viewer response;- how speech or silence carries the piece, who speaks when present, and the roles of A-roll and B-roll;@@ -29,9 +31,10 @@ to overturn the coarse understanding.## Choose evidence for the question-Use `hypit transcribe` for speech and word time. Use `hypit observe` when motion, sound, sequence, or-whole-piece visual context matters. The local `hypit media` family prepares views without interpreting-them:+Use `hypit transcribe` for speech and word time. Use `hypit observe` as the normal broad visual pass+for reference reconstruction, over the whole short work or meaningful spans of a longer or denser+work. Motion, sequence, persistence, transitions and whole-piece context need moving-image evidence.+The local `hypit media` family prepares views without interpreting them:- `probe` reports media facts;- `cut` isolates a short temporal question;@@ -68,6 +71,23 @@ connected whole-piece judgment in `ANALYSIS.md` and locatable facts in `TIMELINEquestion can instead use one clip, grid or frame as its input. The command forms are composable; thecurrent uncertainty decides which of them is useful.+Frames and tiles complement moving-image observation by supplying direct static evidence. If no+moving-image observer is reachable, distinguish the evidence that remains direct from the+relationships that are not yet well established. Frames and tiles still show appearance, text,+composition, geometry, and sampled states; they provide less evidence about motion, transitions,+persistence, and relationships across distant moments. Explain that limitation in terms of the+current work, then use the+[Runtime Profile](../environment/profile.md) to present the actual ways to strengthen the missing+moving-image evidence. A user may knowingly choose to continue with that evidence limit, and a truly+static question may need only static evidence, but the Agent does not present the limited account as+a complete moving-image reading.++When moving-image observation becomes reachable later, use it to revisit the parts of the reference+that were understood from static samples alone. Update `ANALYSIS.md` and `TIMELINE.md` where the+evidence changes them, then reconsider affected Treatment, Script, shots, components, or prompts.+Existing work that the stronger evidence still supports remains useful. Observation may itself be a paid request;+make its Endpoint and price source visible and use only spending authority that covers it.+No evidence form is a substitute for every other one. A still cannot establish movement or sound. Asingle representative frame cannot establish entry, exit, replacement, or persistence. A long videopass can lose small text and rapid order. A tile shows sampled order but may miss what happens betweendiff --git a/skills/hypit/references/environment/model-and-provider.md b/skills/hypit/references/environment/model-and-provider.mdindex bf3d4c3b..5b97a9b0 100644--- a/skills/hypit/references/environment/model-and-provider.md+++ b/skills/hypit/references/environment/model-and-provider.md@@ -50,7 +50,10 @@ concrete request, and an expected result type.The Provider implements that capability through its service. It owns wire fields, URLs, uploads,authentication, task handling, result transfer, service-specific support limits, capacity, diagnostics,-and its price source. Generated media returns as stored Resource references of the declared type.+and its price source. When useful machine-readable pricing exists, `readPricing` may use the complete+request to retrieve relevant Provider-owned material. The returned data keeps the Provider's own shape;+a pricing page remains the interface when that is what the service publishes. Generated media returns+as stored Resource references of the declared type.The shared generated-media vocabulary carries text, media references, and image/video/audio results;it does not choose a vendor or model on the author's behalf.diff --git a/skills/hypit/references/environment/profile.md b/skills/hypit/references/environment/profile.mdindex 0148bfb1..67f1492c 100644--- a/skills/hypit/references/environment/profile.md+++ b/skills/hypit/references/environment/profile.md@@ -156,9 +156,11 @@ Inspect the selected Profile, configured credentials, and current Endpoint diagnthe user to set up anything. For a typical generated reconstruction, assess visual observation,word alignment, image generation, and video generation together so that solving one missingcapability does not conceal another. Keep a working choice. When capabilities are missing, explain-all material consequences in the language of the work and ask once which supported Provider accounts-or Keys the user already has and whether local WhisperX is practical. Present only choices that-materially differ in cost, privacy, setup time, or control.+all material consequences in the language of the work and ask which supported Provider accounts or+Keys the user already has and whether local WhisperX is practical. Present the related gaps together+rather than as serial setup interruptions, and revisit the choice when the user's intention or the+available environment materially changes. Present only choices that materially differ in cost,+privacy, setup time, or control.- Use a supported BYOK Provider when the user already has that account and wants to use it.- For speech alignment, offer local WhisperX when the machine and the user's available setup time@@ -170,6 +172,13 @@ materially differ in cost, privacy, setup time, or control.- Image, video, voice, and audio production each require an actual selected Endpoint for the exactauthored model. An observation Endpoint does not imply a generation Endpoint, and vice versa.+A newly selected Provider can make several capabilities reachable at once. Diagnose the capabilities+it actually offers, then bring every relevant new ability back to the current work rather than using+only the capability that prompted setup. For example, if a choice made for generation also restores+moving-image observation, return that eye to the reference investigation and apply the+[reference-video guidance](../creation/reference-video.md) to the earlier reading. That observation is+still an external request with its own price source and spending authority.+HypiHub is a convenient selected Provider, not an automatic fallback. Moving from BYOK or localexecution to HypiHub changes the Profile or its binding explicitly; an authentication error, exhaustedquota, rate limit or service failure never authorizes that switch or a different billing account.@@ -241,9 +250,11 @@ exist:With no image or video generation account, HyperFrames can still generate deterministic MG, Caption,and Typography when that form fits the requested work; it does not replace generated A-roll or B-roll.-Without reliable visual observation, the Agent can inspect prepared frames and tiles but has weaker-evidence about complete motion and long-range sequence. Without alignment, semantic attachment to real-spoken words is unavailable. State those differences in the language of the requested work.+For reference reconstruction without moving-image observation, use the+[reference-video guidance](../creation/reference-video.md) to explain the evidence limit and what that+means for the work. The user's own Gemini access and the official HypiHub Endpoint are the normal+hosted ways to restore that eye. Without alignment, semantic attachment to real spoken words is+unavailable. State those differences in the language of the requested work.When the user brings another model, service, or Key, use[Models and Providers](model-and-provider.md) to distinguish credential setup, Endpoint configuration,diff --git a/skills/hypit/references/playbooks/craft/compositing.md b/skills/hypit/references/playbooks/craft/compositing.mdindex 49d526c1..3d62ce23 100644--- a/skills/hypit/references/playbooks/craft/compositing.md+++ b/skills/hypit/references/playbooks/craft/compositing.md@@ -4,6 +4,21 @@ Read this when several visual contributions share a frame: a persistent board ova live inset, a cutout, a screen inside a device or a designed background carrying media and text.[Tracks](../../production/tracks.md) owns the authoring mechanics; this page owns the visible relationship.+## Keep physical, camera and editorial space distinct++One finished frame can relate three kinds of space:++| Space | What it owns |+| --- | --- |+| Physical scene | The people, setting and objects actually recorded or generated; gaze, touch and physical interaction belong here. |+| Camera composition | The viewpoint, framing, subject placement, scale and crop through which that scene is seen. |+| Editorial composite | The Caption, MG, inset, mask and other Tracks placed over or around the source picture, including their stack order. |++The final design can make these spaces answer one another without merging them. A later overlay can+influence the camera composition chosen for a source picture, but it does not become an object inside+that picture's physical scene. The source remains a complete image of its world; the overlay remains+an authored layer in the final composition.+## Give the layers room togetherDecide which information leads at each beat. A small inset can still dominate through movement,@@ -11,15 +26,16 @@ contrast or a face; a large quiet field can support the main action. Keep necessgestures, product labels and readable text clear of persistent graphics.Plan composition before generating the base image when a long-lived overlay reserves part of it.-Describe a natural scene that leaves the useful area free, such as a person to the right with a table-on the left. Keep the intended graphic out of that scene prompt when it will be authored separately.-The smallest owning response to a collision may be a new crop, a moved overlay, simpler content or a-newly framed picture.+Frame the subject and anything that must remain readable outside the overlay's planned footprint while+directing the whole image as a complete natural scene. Keep the intended graphic out of that scene+prompt when it will be authored separately. The smallest owning response to a collision may be a new+crop, a moved overlay, simpler content or a newly framed picture.Canvas dimensions, intrinsic media extent, destination Frame and fit/crop are separate facts. Checkthe result of all of them, not just the Frame coordinates. A tall cutout can consume most of avertical Canvas even when its width seems modest. Stack order is explicit inside Tracks, not the-order they happen to appear in Film.+order they happen to appear in Film. [Spatial layout](../../production/spatial.md) owns the exact+Frames and fitting once the intended relationship is known.## Distinguish container motion from content motion@@ -49,17 +65,49 @@ Inspect at delivery size. Glare, tiny lettering, crop, moving fingers or captionfact the insert was meant to show. A plausible-looking product interaction is not evidence of a realproduct capability by itself.+## Choose an A-roll presentation++The performance carrying a spoken Segment remains A-roll across several common visual+presentations. Choose the source shape and final Frame from the intended relationship, the reference+evidence, and the feeling of the shot rather than from a fixed layout recipe.++| Presentation | Useful starting judgment |+| --- | --- |+| Full-frame | A `9:16` character-and-scene image and reference-generated performance are the common fit for vertical creator footage. Frame the face, body and real setting for the later composite. |+| Upper or lower split | A square camera image and performance often cover a `9:8` half of a vertical Canvas cleanly, with a small top/bottom crop. A close, intimate phone view may instead be a vertically generated performance whose other half is covered. |+| Foreground cutout | A vertical performance often preserves a natural amount of the speaking body. Remove its background, normalize the alpha-bearing result, and use that same prepared performance as the SemanticTake. |+| Circular picture-in-picture | A square source with the face near its own center is a useful starting point. Give Speech Track a square Frame and rounded clip whose radius is half the side; its semantic, visual and audio outputs remain one assembly. |+| Audio-only A-roll | The performance contributes semantic time and sound but no picture. Another Track fills the Segment with the intended demonstration, POV, B-roll, Typography or MG. |++For a non-overlapping upper/lower split, equal stacking order can express two peer regions; the+Frames themselves keep them separate. Other unambiguous stack choices can serve the same picture.+Use Content Fit to map the source into its destination Frame rather than changing the source's+semantic role.++When reconstructing an apparent half-screen presenter, the final rectangle alone does not reveal its+source shape. Inspect camera distance, intimacy, body crop, background coverage and related views. A+close phone feeling can support the reading that a vertical performance is partly covered; a flatter,+more distant composition can support a native square view. These are useful priors, while the actual+visual evidence and intended target composition decide.++The source image and video direction describe a complete physical scene and camera view. Position a+face high or low, left or right, through framing and meaningful scene relationships when that serves+the final layout. Editorial graphics remain separate layers and are not objects for the generated+person to see, touch or point toward. [Image direction](image-direction.md#frame-the-image-for-what-it-will-become)+owns the camera-image wording; [Spatial layout](../../production/spatial.md) owns Frame, fit and crop.+## Inset, cutout and flattened composite are distinct choices-A rectangular image or normalized video can be placed directly as a Media Item. A circular inset-is a geometric crop: use a square Frame and a rounded clip whose radius is half that square's side,-or an authored clip Path. The camera background remains inside the circle. A cutout instead follows-the person's silhouette and needs suitable transparency in the source.+A speaking rectangular inset, circular inset or normalized cutout can be presented directly by+Speech Track. A circular inset is a geometric crop: use a square Frame and a rounded clip whose+radius is half that square's side. The camera background remains inside the circle. A cutout instead+follows the person's silhouette and needs suitable transparency in the source. Speech Track presents+the Take while it carries the Segment's speaking performance. Media Track can use the same prepared+media when the work intentionally gives it an independent Window, playback, trim or visual role.A speaking person in either presentation remains A-roll when their performance carries the main-Script. A large screen recording can occupy the background while a small presenter above it supplies-the semantic timeline. Choose placement and stacking for the viewer's attention; preserve the-performance's semantic timing and route its speech once.+Script. Choose placement and stacking for the viewer's attention; preserve the performance's+semantic timing and route its speech once.When the intended composition needs an isolated silhouette, a still portrait can use[image background removal](../../production/image-operations.md#remove-a-background). A moving subject@@ -89,9 +137,9 @@ require another kind of Take or Track. Choose which role the cutout serves in th| A speaking performance establishing semantic time | `cutout-media.media` → SemanticTake → Speech Track; include its visual and audio outputs in Film. || A visual overlay on an existing semantic timeline | `cutout-media.media` → a Media Item's `media` input; select its Window and include the Track's visual output in Film. |-A deterministic image composition can flatten several layers into a reusable still when that frozen-result is the useful artifact. If a cutout belongs inside it, removal precedes composition. An ordinary-inset or a source with suitable transparency already has the material it needs.+When the intended artifact is a fixed flattened arrangement of still images, use+[Image Compose](../../production/image-operations.md#flatten-a-fixed-still-image-arrangement).+That page owns its exact role, limits and Source syntax.Inspect cutout edges against the final background: hair, fingers, translucent edges, holes, halosand color spill. Match light and color when the layers should read as one scene; an intentionallydiff --git a/skills/hypit/references/playbooks/craft/image-direction.md b/skills/hypit/references/playbooks/craft/image-direction.mdindex c83a7e4f..a41f1b2a 100644--- a/skills/hypit/references/playbooks/craft/image-direction.md+++ b/skills/hypit/references/playbooks/craft/image-direction.md@@ -164,7 +164,7 @@ Choose a small number of details that carry that relationship: condensation on aKorean-style café, a product diagram on a startup whiteboard, a cat tree that is visibly too small foran orange-cat CEO, drying flowers on a textured wall, or worn candles inside a real stone chapel.Favor details that connect choices across the image. The record-store owner's turntable belongs to-her work and gives her rightward placement a spatial reason; a handwritten recommendation card+her work and gives the composition a physical counterweight; a handwritten recommendation cardexpresses how she runs the shop. Make the details that define this person, place, and shot explicit,then leave incidental particulars open for the model to resolve within that direction.@@ -221,12 +221,18 @@ should be readable while the shoulders and upper body still establish the personintensifying “close” can instead favor a face-dominated crop. Asking to keep the body in frame meanspreserving the intended visible silhouette, not widening the shot until the whole body fits.-Choose left, center, or right placement from the scene and the eventual composition. A slight offset-can give the image a casually captured quality, but it should have a spatial reason. A small table-entering the lower-left corner can balance a person on the right; a chair or thighs angled toward-the left can establish the same relationship through posture. Describe where the person sits in the-place, rather than merely moving a centered portrait toward an edge. The balancing area remains a-real part of the setting, with its own objects and texture.+Choose left, center, or right placement from the intended camera view and eventual composition.+Off-center placement can be a direct camera decision; it needs no added prop to explain it. When a+table, chair, partner or product truly belongs in the generated scene, its geometry can also motivate+the person's posture, gaze or contact and contribute balance inside the frame. Choose those objects+for the physical scene's own story, while camera composition reserves room needed by the editorial+layout. Describe the person and setting as one actual camera image.++Give gaze, posture, touch, and other interaction only to people or objects that actually belong in+the generated image. A later Caption, icon, product card, or other MG can influence framing and the+subject's placement, but it is not an object in the character's physical world. Keep the person+engaged with the camera or with the real scene, and direct the generated setting as the complete,+natural place it should be on its own.Use seat and leg details when they explain that relationship. A brief direction such as “her thighsangle toward the table on her left” can be useful; a catalog of shoes or clothing outside the intended@@ -237,18 +243,21 @@ Apply that speaking-reference solution when the work has the same need. A side-oproduct action, or full-body shot needs its own gaze, hand occupancy, contact, and composition.A specific starting action is useful when the shot is actually about that action.-Compose for the final stack. A Ranking icon on the left can call for a presenter toward the right;-the same person may remain on that side across every related shot. Keep the face, microphone, and-useful gesture area clear of the overlay's actual footprint and movement. A lower-left table can-balance the shot while leaving an upper-left icon area available. Position and scale depend on the-work, not a fixed left/right alternation or a universal percentage.--Translate that layout into visible image direction: describe subject placement and usable background-space. Author the MG separately in its component; mentioning a future icon is not a request to paint-it into the source image. Carry the intended spatial relationship into the downstream performance.-If Caption follows a face, leave usable space around the head. If a board covers the lower frame,-place the performer in the region that remains visible. If a product or screen must be read, give-it enough scale and a credible relationship to the person.+Compose for the final stack. A Ranking icon planned for the left can call for a presenter toward the+right while the presenter continues speaking to the camera or acting inside the real scene. The same+person may remain on that side across every related shot. Keep the face, microphone, and useful+gesture area clear of the overlay's actual footprint and movement. The rest of the source image still+belongs to its real setting. Position and scale follow this work's visual hierarchy. [Compositing](compositing.md)+distinguishes the generated scene, camera composition and later editorial layers.++Translate that layout into visible image direction: describe subject placement, camera framing, and+the real setting. Author the MG separately in its component. The image prompt describes the pixels+this request should produce; downstream editorial content stays in the downstream composition. Carry+the intended placement into the generated performance without inventing an interaction with that+later layer. Frame the face at a height compatible with a face-following Caption. If a board covers+the lower frame, place the performer in the region that remains visible. If a product or screen is+part of the generated scene and must be read, give it enough scale and a credible relationship to the+person.## Make imagined subjects physically presentdiff --git a/skills/hypit/references/playbooks/craft/seedance-directing.md b/skills/hypit/references/playbooks/craft/seedance-directing.mdindex c5b3dab4..7e99bbfd 100644--- a/skills/hypit/references/playbooks/craft/seedance-directing.md+++ b/skills/hypit/references/playbooks/craft/seedance-directing.md@@ -65,6 +65,12 @@ A speaking Take can be assembled directly from the installed Source:The selected model vocabulary supplies its current duration, reference-count, format, and medialimits. The Kit does not own those limits or the request's explicit media edges.+For this visible A-roll form, the ReferenceVideo request is the speaking performance: it receives+the short Voice Reference and exact Script dialogue, then returns the person's performed picture and+sound together. Voice Design establishes who that person sounds like, so its short casting line can+be unrelated to the finished dialogue. Voice Clone instead produces audio-only A-roll when a passage+has no on-camera speaking performance.+Use the Script's `.dialogue` Text rather than retyping dialogue into action. Map Role names to theKit's A/B identities where necessary. Preserve one recurring voice reference per character whenvoice continuity matters. See [Voice and performance](voice-and-performance.md) for A-roll versus@@ -78,6 +84,18 @@ open-palmed explanation, a knowing shrug or a short glance can carry more than eVibe alone may leave an important interaction unspecified; describe the decisive action when thestory needs a handoff, a look toward a partner or an exit.+The action prompt describes what this generation should make visible over time. Give the performer+physical relationships only with the camera, people, props, and parts of the setting that actually+exist in the generated scene. A Caption, icon, product card, or other MG that will be composed later+can shape where the performer is framed, but it is not something for the performer to point toward,+touch, watch, or reveal inside this Take.++Capable video models can turn figurative wording into literal objects, events, or transformations.+Use concrete visible language for the intended gaze, gesture, movement, camera behavior, and cuts+when a metaphor would introduce the wrong scene content. Social attitude and aesthetic shorthand+remain useful when they direct the performance; an imagined object or event belongs in the prompt+only when it should truly appear in the generated video.+Natural emphatic gestures are usually more reliable than asking fingers to display an exact number.Let speech, Caption or MG convey the quantity while pointing and hand actions serve the performance.diff --git a/skills/hypit/references/playbooks/craft/voice-and-performance.md b/skills/hypit/references/playbooks/craft/voice-and-performance.mdindex fb2ed824..b73004f2 100644--- a/skills/hypit/references/playbooks/craft/voice-and-performance.md+++ b/skills/hypit/references/playbooks/craft/voice-and-performance.md@@ -3,24 +3,68 @@Read this when deciding who performs spoken material, how a recurring person keeps one voice, orwhether an off-screen passage continues an A-roll performance or uses independent speech.-## Begin with who carries the expression--Most presenter-led short videos get their speech from an A-roll performance. The generated Take-carries the words, mouth movement, gesture, gaze, delivery, and picture together. An accepted camera-image establishes the person in a useful shot, an accepted Voice Reference establishes how they-sound, and the video direction tells the performer what to say and do.--A-roll is the performance carrying the main Script and establishing its semantic timeline. Its-SemanticTake retains that role when the picture becomes a circular inset, a moving cutout over a-demonstration, or a full-frame presenter. The same performance remains A-roll when temporarily-covered by B-roll or MG. The viewer may stop seeing the speaker while continuing to hear the same performance.-The covering B-roll may even show the same person doing a silent lifestyle action. Seeing a person while-hearing words does not by itself identify that picture as their speaking Take or create a new speech source.--Read the whole work or passage before choosing the source of speech. A recurring presenter, podcast-speaker, interview participant, or host normally continues to carry their own lines through visual-coverage. A passage built as narration from the outset may instead use independent speech while-hands, screens, products, montage, B-roll, Typography, or MG carry the picture.+## Let the Segment reveal the A-roll++A Segment is one authored passage in the Script. Before production it has identity and meaning but no+seconds. In Hypit, A-roll names the performance that makes a spoken Segment time-bearing. It can be+a visible A-roll whose picture and sound are performed together, or an audio-only A-roll whose+independent speech carries the passage while other Tracks supply the picture.++The A-roll does not fit itself onto an earlier program timeline. By performing the Segment, it gives+that passage real time; the other Segment performances together create the program's semantic time.+[Tracks](../../production/tracks.md#semantic-takes-supply-the-media-timeline) owns the exact+SemanticTake assembly and projection mechanism.++This gives a practical question: **who is speaking this Segment?** For visible A-roll, that person's+generated Take carries words, mouth movement, gesture, gaze, delivery, picture, and sound together.+In a multi-person exchange, the speaking performance can include listeners, several turns, or camera+cuts while still carrying one Segment. The number of people visible in the frame does not determine+how many semantic clocks exist.++A recurring presenter, podcast speaker, interview participant, or dramatic character continues to+own their own lines even while another picture covers them. Visual handoffs, coverage, Caption, MG,+and Effects are authored against the time that performance established. If an edit changes the+performance's actual length, prepare and align the edited media before assembly.++A wordless Segment is the important edge of the same model. Nobody speaks, so it has no speaking+A-roll, but it still has authored meaning. Its prepared media supplies the Segment's start and end,+and its SemanticTake carries an empty word array into the same SemanticTrack. Here semantic means the+authored identity and relationships of the passage, not a requirement that it contain speech.++## Keep the performance role separate from the picture++A-roll is not the bottom layer, the largest rectangle, or the currently visible face. Its+SemanticTake retains the role when its picture is full-frame, one half of a split, a circular inset,+a moving cutout, or completely covered by B-roll or MG. A large app recording can be the primary+picture while a small presenter in a corner supplies the semantic timeline and speech.++The covering B-roll may even show the same person doing a silent lifestyle action. Seeing a person+while hearing words does not by itself make that picture the speaking Take or create a new speech+source. More fundamentally, the sound already belongs to the A-roll performance that created this+Segment's semantic time. B-roll changes or supplements the visual answer; it does not inherit the+speech merely because a person appears in it. Follow that positive ownership relationship through+the composition. [Compositing](compositing.md#choose-an-a-roll-presentation) owns the full-frame,+split, cutout and circular presentation choices.++## Let neighboring A-roll performances keep their time++The usual A-roll assembly places Segment performances one after another without consuming either+side of their duration. This fits how the work is authored: `hypit measure` helps size each request so+its words, delivery and action belong naturally in that generated passage, and speaking video models+normally use the requested clip to perform the line. Asking one Take to remain silent for a synthetic+half-second handle at its beginning or end works against both that sizing and the model's performance.++A duration-consuming crossfade between neighboring A-roll Takes can eat into the word windows that+each performance established. The result may contain simultaneous speech at a boundary and weaken+the listening rhythm. For ordinary creator speech, podcast turns, interviews and short drama, this+makes a direct join between the prepared Segment Takes a practical starting point. Visual coverage+and effects can still cross that seam without changing the underlying speech time.++This is a directing judgment, not a ban on another temporal design. Deliberate interruption,+overlapping dialogue, musical phrasing or another work may call for a different assembly. A project+component can own that operation and publish the explicit semantic, visual and audio outputs its+consumers need. [Track authoring](../../production/track-authoring.md#author-an-alternative-performance-assembly)+explains the type boundary; the default Speech Track remains the simple lossless assembly.## Give a recurring person one accepted voice@@ -28,6 +72,10 @@ A **Voice Reference** is an ordinary accepted audio Resource that establishes aidentity. Reusing that Resource wherever the same person performs is what carries the relationshipthrough the production.+Give each recurring person who speaks their own Voice Reference. A silent listener does not acquire+a voice merely by appearing in the picture; when that character speaks elsewhere, their own+reference follows them into that performance.+When the user supplies the exact private voice they want, prepare a clean representative excerpt asthe Voice Reference. Otherwise direct Voice Design from the person's role, energy, age range,language, accent, vocal texture, pace, and delivery. Voice Design acts as a broadly capable casting@@ -52,31 +100,83 @@ Voice Design or supplied audio↓accepted Voice Reference↙ ↘-A-roll performance independent speech+visible A-roll audio-only A-roll```An A-roll-capable video model receives the reference while generating the person's visible speakingTake. Voice Clone receives the same reference when the work genuinely needs that person to speak asan independent audio source. These are two uses of one ordinary Resource, not two voice identities.-A work may combine them. A host can perform the main A-roll and later narrate a passage that has no+A work may combine them. A host can perform visible A-roll and later narrate a passage that has nounderlying on-camera performance; using the same Voice Reference makes both sound like the sameperson. Another passage covered by B-roll may still be audio from the A-roll Take beneath it. Decidefrom the passage's expressive construction, not from whether the face happens to be visible at thatinstant.+## Choose between visible and audio-only A-roll++For visible A-roll, the video request receives the character's camera image, Voice Reference and+exact Script dialogue, then returns the picture and sound of that person actually performing. Voice+Design casts the person; its short sample line exists to establish voice identity and need not repeat+any line from the finished Script.++Audio-only A-roll is useful when a Segment is constructed without an on-camera speaking+performance: a desktop point-of-view demonstration, a narration-led montage, a pure MG explanation,+or a product-explanation passage inside an otherwise presenter-led work. Voice Clone+uses the person's existing Voice Reference and the Segment's actual Script to produce that passage.+The resulting audio-only SemanticTake carries the A-roll role even though it contributes no picture;+Media Track, Typography, MG, or another visual Track supplies what the viewer sees.++Choose audio-only A-roll because the passage itself is independently narrated, not merely because+B-roll or graphics happen to hide a visible performance. Use Voice Design to establish the person's+reusable voice and Voice Clone to perform the actual Segment; the casting sample and the finished+delivery have different jobs.++Use `hypit vocabulary @hypit/mimo-speech` for the installed Voice Design and Voice Clone author+Surfaces. The Source chooses those model semantics; the Runtime Profile independently chooses the+Endpoint that can perform them.+## Treat narration as a sound-picture relationshipNarration means the audience hears speech while the speaker is not the primary visible action. Thatdescription does not reveal how the speech was produced.-- In an A-roll-led passage, coverage changes the picture while the accepted speaking Take remains the- performance and timing authority.-- In a narration-led passage, accepted independent speech supplies the performance while the picture- is designed separately around it.+- In a visible-performance-led passage, coverage changes the picture while the accepted visible+ A-roll remains the performance and timing authority.+- In an independent-narration-led passage, audio-only A-roll supplies the performance while the+ picture is designed separately around it.- In a mixed work, each passage can use the relationship that serves it, while recurring people keeptheir accepted Voice References.+Keep three questions distinct:++| Question | What it establishes |+| --- | --- |+| Is the speaker off screen at this instant? | A viewing relationship: speech is heard while its performer is not the primary visible action. |+| Does the sound come from covered visible A-roll or from audio-only A-roll? | The production source and owner of the performed audio. |+| Is this passage visible-performance-led or independent-narration-led? | The Segment's higher-level expressive construction. |++The same visual language can answer these questions differently. Consider a notebook-paper field+with animated steps, labels and panels:++- When that field persists as the place where a product process is explained, its changing graphic+ relationships carry the passage. Independent speech can establish the Segment's semantic time+ while the paper composition supplies its picture. This is an independent-narration-led module.+- When a ranking host says a brief “Top one: the product” over a paper reveal and then appears on+ camera to continue the same judgment, the host's performance normally owns the whole passage. The+ paper reveal is MG coverage over that A-roll, and the sound remains the host Take's sound.++Duration can support the reading but does not decide it. Ask what carries the argument, attitude and+progression, and whether an on-camera performance exists whose words naturally continue through the+coverage. A sustained graphic explanation and a brief graphic reveal can look alike in one frame+while belonging to different constructions.++Segment boundaries express the work's useful modules without acting as source labels. One+independent-narration-led Segment can contain many MG steps through Selections and Moments. One+visible-performance-led Segment can begin under a graphic and later reveal the speaker. Splitting a+generation into two Segments also does not make either one audio-only A-roll by itself; each+Segment's actual performance relationship remains the deciding fact.+Use `../formats/narration-led-demo.md` when independent narration actually organizes a work orsubstantial passage. `../index.md` routes to available Format and Craft knowledge for the work'sother relationships.diff --git a/skills/hypit/references/playbooks/formats/narration-led-demo.md b/skills/hypit/references/playbooks/formats/narration-led-demo.mdindex c7e75bed..49b1c4fd 100644--- a/skills/hypit/references/playbooks/formats/narration-led-demo.md+++ b/skills/hypit/references/playbooks/formats/narration-led-demo.md@@ -5,15 +5,15 @@ hands, screens, demonstrations, B-roll, Typography, or MG carry the picture.## Recognize the organizing relationship-A narration-led passage is organized by an accepted independent speech performance. Its pictures-are designed around that performance rather than being temporary coverage over a speaking A-roll-Take. Common forms include desk demonstrations, screen tutorials, product walkthroughs, visual-essays, montage narration, and MG explainers.+A narration-led passage is organized by accepted independent speech: audio-only A-roll. Its pictures+are designed around that performance rather than being temporary coverage over visible A-roll.+Common forms include desk demonstrations, screen tutorials, product walkthroughs, visual essays,+montage narration, and MG explainers.The decisive question is not whether a face is currently visible. A presenter-led work may cover its-A-roll with B-roll for several seconds and still be A-roll-led. A narration-led passage begins with a-speech performance that does not depend on an underlying on-camera delivery. A mixed work can move-between these relationships when the story does.+visible A-roll with B-roll for several seconds and remain visible-performance-led. An+independent-narration-led passage begins with audio-only A-roll that does not depend on an underlying+on-camera delivery. A mixed work can move between these relationships when the story does.State that relationship in Treatment. It determines how performance, picture coverage, visualproof, and later review should be understood.@@ -55,9 +55,9 @@ and README for its exact inputs and limits.## Keep the narrator one person across the workUse `../craft/voice-and-performance.md` to establish the narrator's accepted Voice Reference and-generate independent speech from it. If the same person also appears in A-roll elsewhere, the same-Voice Reference can condition both the visible performance and independent narration. The work then-changes sound-picture relationship without changing who is speaking.+generate independent speech from it. If the same person also appears in visible A-roll elsewhere,+the same Voice Reference can condition both the visible performance and independent narration. The+work then changes sound-picture relationship without changing who is speaking.`../../creation/script-and-time.md` owns measurement and alignment. One accepted narrationperformance can carry many picture changes; their semantic relations follow that performance withoutdiff --git a/skills/hypit/references/playbooks/index.md b/skills/hypit/references/playbooks/index.mdindex de0a92ee..99cbe2b5 100644--- a/skills/hypit/references/playbooks/index.md+++ b/skills/hypit/references/playbooks/index.md@@ -23,7 +23,7 @@ not need a matching playbook before it can be understood or made.| Read | When the current question concerns || --- | --- || [Directing generated images](craft/image-direction.md) | phone-video realism, character presence, image prompts, materials and palette, or preparing a useful visual reference |-| [Voice and performance](craft/voice-and-performance.md) | who performs speech, recurring voice identity, A-roll under coverage, or genuinely independent narration |+| [Voice and performance](craft/voice-and-performance.md) | how Segment performances create semantic time, who is A-roll, recurring voice identity, visual A-roll presentations, or genuinely independent narration || [Directing generated performance](craft/seedance-directing.md) | using a video Kit, directing attitude and small actions, natural cuts, listeners, or sizing Takes || [Reference relationships](craft/generated-dependencies.md) | deciding which images or Takes should depend on which references || [Visual continuity](craft/visual-continuity.md) | complementary camera views, gaze, prop state, or what must remain coherent through a cut |@@ -33,12 +33,19 @@ not need a matching playbook before it can be understood or made.| [Caption tracking](craft/caption-tracking.md) | measuring produced footage and placing Caption above a moving head without regenerating the media || [Sound and mix](craft/sound-mix.md) | the relationship among speech, music, ambience, effects, loudness, ducking, and continuity || [Graphic composition](craft/graphic-compositions.md) | visual hierarchy, boards, graphic state, editable structure, component boundaries or coordinated palette and reveals |-| [Compositing](craft/compositing.md) | overlays, picture-in-picture, cutouts, screen evidence, container versus content motion, or live call layouts |+| [Compositing](craft/compositing.md) | physical scene versus camera framing and editorial layers, overlays, picture-in-picture, cutouts, screen evidence, or live call layouts |Read every craft whose question genuinely appears in the work, including questions discovered aftera Studio preview or completed Result. Repeated use of the same craft does not require rereading itunless the decision has changed.+## Examples++| Read | When a concrete directing example would help |+| --- | --- |+| [Image direction examples](craft/examples/image-direction.md) | complete phone-video prompts, their visible results, and the aesthetic relationships behind them |+| [Conversation image examples](craft/examples/conversation-images.md) | complementary podcast and interview views, product references, lifestyle B-roll, and changing camera relationships |+## Formats| Read | When the work's shape includes |diff --git a/skills/hypit/references/production/builds.md b/skills/hypit/references/production/builds.mdindex c4361035..f4107d63 100644--- a/skills/hypit/references/production/builds.md+++ b/skills/hypit/references/production/builds.md@@ -70,9 +70,27 @@ would be sent. Planning starts no external work.With a selected Runtime Profile, the plan also names the Endpoint behind each request, applies thatEndpoint's request-support rules, checks cheap readiness for the demanded capabilities, and shows the-Provider's declared price page or that its price source is unknown. Hypit does not copy changing rate-tables or manufacture a total. Use the linked Provider source together with the plan's real request-counts and authored parameters when a cost judgment is needed.+Provider's declared price page or that its price source is unknown. Planning remains local and does+not require a billing account to answer.++Read the current pricing material selected Providers expose for these same Needs:++```bash+hypit pricing path/to/build.svrun+```++`pricing` is an explicit read-only network operation. It creates no Build and submits no generation.+It places each Need beside the current documents or page supplied by its selected Endpoint. Providers+may expose a narrow model record or a broader catalogue; Hypit does not impose a common price-table+shape or calculate a total. Use the Need's authored duration, resolution, count, or other billing facts+to calculate and explain the relevant cost. Missing or broad pricing material is uncertainty to report,+not a system prohibition on an otherwise authorized Build.++The default human view groups matching Needs and keeps Provider documents folded to their source URLs.+Use `--json` when you need the original documents for calculation, or `--verbose` to inspect them in the+terminal. A mixed production naturally has separate image, video, voice, alignment, and local-processing+entries. Read each relevant entry against its own Need, then explain the production's expected cost in+ordinary language.Read the remaining Needs against this change. A Caption or MG-only revision should keep its existingmedia generation satisfied; replacing selected B-roll should leave the unchanged performance satisfied.@@ -80,8 +98,8 @@ Rendering and other required processing may still appear as Needs. Explain eachfrom the user's goal or an explicit generation decision. If a request appears because a Candidate waslost or never selected, repair the Run and plan again before asking to spend or submitting work.-Before new paid work, tell the user which requests and Provider Endpoints the plan selected and link-their price sources. Existing explicit authorization for that described work is sufficient; otherwise+Before new paid work, tell the user which requests and Provider Endpoints the plan selected and report+the available calculation or link its Provider source. Existing explicit authorization for that described work is sufficient; otherwiseobtain authorization before `build`. Authorization to correct downstream work does not silently extendto regenerating unchanged media. Additional paid work outside the authorized scope needs a new decision.diff --git a/skills/hypit/references/production/image-operations.md b/skills/hypit/references/production/image-operations.mdindex a5cd64ec..a96d314e 100644--- a/skills/hypit/references/production/image-operations.md+++ b/skills/hypit/references/production/image-operations.md@@ -8,22 +8,39 @@ An image operation publishes another ordinary image Output. The resulting imagea Media Track, another image operation, or any compatible project component. The operation belongsin Source when that transformation is part of the reproducible production relationship.-## Compose several images on one canvas+## Flatten a fixed still-image arrangement-Use Image Compose when the work needs a flattened picture made from explicitly positioned layers:+Image Compose has one narrow job: preserve a deliberately rigid two-dimensional arrangement of+existing still images and publish it as one PNG. A comparison board, contact sheet or intentionally+hard-edged collage can need exactly that operation.++Declare the Canvas and every rectangular destination explicitly:```svml+<import as="space" from="@hypit/spatial@1"/><import as="compose" from="@hypit/image-compose@1"/>-<compose:Image id="card" canvas={portrait} background="#00000000">- <compose:Layer source={background.image} frame={full} fit="cover"/>- <compose:Layer source={product.image} frame={product-frame} fit="contain"/>+<space:Canvas id="comparison-canvas" width="2048" height="1024"/>+<space:Frame id="before-panel" within={comparison-canvas}+ left="0%" top="0%" right="50%" bottom="100%"/>+<space:Frame id="after-panel" within={comparison-canvas}+ left="50%" top="0%" right="100%" bottom="100%"/>++<compose:Image id="comparison" canvas={comparison-canvas} background="#EEEAE2FF">+ <compose:Layer source={before.image} frame={before-panel} fit="contain"/>+ <compose:Layer source={after.image} frame={after-panel} fit="contain"/></compose:Image>```Child order is paint order. Each Layer keeps the source image, destination Frame, fit and optional-opacity explicit. `{card.image}` is a flattened image, not a Track; use a visual component instead-when the layers need independent timing or motion in the final video.+opacity explicit. `fit` accepts `contain`, `cover` and `stretch`; `interpolation` selects the raster+resampling filter. The background is `#RRGGBBAA`, so `#00000000` is transparent. The published+`{comparison.image}` is one flattened PNG rather than a Track.++Image Compose does not understand people, objects or camera space. It cannot reconcile perspective,+depth, lighting, subject scale, background continuity or a natural seam. When several references+should become one coherent camera image, direct the image model to create that image. When the layers+must remain independently timed, clipped, moved or revised in the video, keep them in Tracks and Film.## Apply a reusable correction programdiff --git a/skills/hypit/references/production/media.md b/skills/hypit/references/production/media.mdindex 9bac3392..84f034e5 100644--- a/skills/hypit/references/production/media.md+++ b/skills/hypit/references/production/media.md@@ -135,6 +135,12 @@ is assigned by the downstream Source relationships just like any other video. Adframe count and progress ruler; the guide is one display treatment and says nothing about later Filmplacement.+StillVideo itself is not a preview state. A held image intended to remain in the Film is ordinary+production media. When a Run uses a held image only in place of a future moving shot, that Candidate+is temporary: keep the actual image as evidence while it is useful, then select the produced video+when the question depends on motion, performance or changing overlap.+[Runs](runs.md#use-the-most-representative-visual-evidence-available) owns that choice.+## Edit bytes at an explicit point in the graph```svmldiff --git a/skills/hypit/references/production/review.md b/skills/hypit/references/production/review.mdindex 8fd1e295..5011eb2e 100644--- a/skills/hypit/references/production/review.md+++ b/skills/hypit/references/production/review.md@@ -12,6 +12,12 @@ Open Studio with the ordinary Run intended for the work. That Run binds the Authfiles, earlier Results, and stand-in Candidates, so Studio and a Build see the same authored graph.A separate review Run is useful when its Candidate selection is intentionally different.+Judge a visual fact from the most representative evidence that actually exists. Supplied and produced+images and videos are the primary evidence for their appearance. A held production image can prove+what that image contains and how it composes with deterministic graphics; it cannot prove the future+shot's movement or performance. A produced video is the evidence for its changing face position,+gesture, timing and interaction with Caption or MG.+Studio renders the selected display closure with deterministic Producers and the exact capabilitiesthat the Runtime Provider declares safe for transient authoring. It submits no Build and no paidgeneration. When that closure still needs external generation, use an existing file or Build@@ -21,6 +27,8 @@ suitable pixels exist and a component or wiring question still deserves an earlyA Card can establish that the downstream graph renders, show broad geometry, and carry provisionalclip time. Its visible marking keeps that evidence distinct from subject identity, performance,shot composition, texture, continuity, or the interaction between production media and graphics.+It stops being useful when the same question can be answered from actual production pixels; earlier+use does not create an approval state that the work must preserve.## Separate legality from judgmentdiff --git a/skills/hypit/references/production/runs.md b/skills/hypit/references/production/runs.mdindex fe1f74dc..53b4e844 100644--- a/skills/hypit/references/production/runs.md+++ b/skills/hypit/references/production/runs.md@@ -88,11 +88,37 @@ The resulting Candidate is named `card.video`. Selecting it replaces the performretaining the preparation and composition that consume that video. A substitute can also consumean existing computed Output; that Output remains part of the selected dependencies.-When a usable image already exists, hold that image with `media-pipeline:still-video` or-`media-pipeline:clip-time-still-video` instead. When usable video exists, select that video itself.-The generic Card is useful only while no suitable pixels exist and the current component or wiring-question is worth answering before generation completes; its role naturally disappears as real-production media becomes available.+## Use the most representative visual evidence available++Supplied and produced images and videos are the primary visual evidence for a production. Select a+usable video itself when it exists. While an intended video is still unavailable, an actual image+from the work can be held with `media-pipeline:still-video` or+`media-pipeline:clip-time-still-video`; it carries real subject, color and camera evidence into the+composition, although it cannot establish the future motion or performance. A generic Card is useful+only when no suitable pixels exist and a current component or wiring question is worth answering+before generation completes.++For example, keep a previously produced presenter image, give it provisional clip time, and use that+video Candidate at the position where the future performance will enter:++```svrun+<import as="media" from="@hypit/media-pipeline@1"/>+<build-record id="presenter-image" build="bld_..." output="presenter.image"/>+<satisfy output="presenter.image" candidate="presenter-image"/>++<fragment id="held-presenter" using="media:clip-time-still-video">+ <input name="duration" value="5"/>+ <input name="clock" from="clock"/>+ <input name="source" from="presenter.image"/>+</fragment>+<satisfy output="performance.video" candidate="held-presenter.video"/>+```++Preview choices answer the question that exists now; they are not a sequence of production states.+A production may skip them entirely, use one while media is being generated, or replace one as soon+as more representative media is available. The Card normally loses its purpose first. A held image+used in place of future motion loses that purpose when the video exists. Run Candidate selection+makes each choice explicit without inventing an acceptance or promotion workflow.## A complete layout preview@@ -137,9 +163,10 @@ selected preview media through that environment and reuse its completed Outputs.production Run still requests generation and measured alignment, under the work's spending authority.For an existing production with usable footage, select its Result or file directly. If only an-authored or generated picture exists, StillVideo already gives downstream composition much more-relevant pixels than a Card. The actual shot is needed to judge overlap with a face, color-relationships, performance or identity. [Studio](studio.md) explains what the preview can show.+authored or generated picture exists, StillVideo gives downstream composition more relevant pixels+than a Card. The held picture can establish its own identity, color and framing; the actual shot is+the evidence for motion, performance and the changing relationship between a face and graphics.+[Studio](studio.md) explains what the selected preview can show.## Preserve the choices that still applydiff --git a/skills/hypit/references/production/spatial.md b/skills/hypit/references/production/spatial.mdindex 68804f4e..6524fb9f 100644--- a/skills/hypit/references/production/spatial.md+++ b/skills/hypit/references/production/spatial.md@@ -1,7 +1,10 @@# Spatial layoutRead this when positioning or fitting media, text or a project component. Timing is explained in-[Script and time](../creation/script-and-time.md); it is independent of these coordinates.+[Script and time](../creation/script-and-time.md); it is independent of these coordinates. Read+[Compositing](../playbooks/craft/compositing.md#keep-physical-camera-and-editorial-space-distinct)+when deciding the relationship among the generated scene, camera framing and later visual layers;+this page owns their exact authoring geometry.## Canvas, extent and destinationdiff --git a/skills/hypit/references/production/studio.md b/skills/hypit/references/production/studio.mdindex 702b5014..7419b5d5 100644--- a/skills/hypit/references/production/studio.md+++ b/skills/hypit/references/production/studio.md@@ -49,8 +49,14 @@ only those exact immediate capabilities a Provider explicitly permits for transiIf new image/video generation blocks the display closure, select an existing Output or file; usean existing image through StillVideo when one is available. An explicit Card Candidate can keep acomponent or wiring study moving when no suitable pixels exist yet. It supplies a visible canvas-and provisional clip time; the eventual media establishes performance, identity, composition and-the real relationship between subject and graphics.+and provisional clip time. The selected production image establishes its actual subject, color and+framing; the produced video adds the evidence for performance, motion and changing overlap with+graphics.++The Preview view renders what the Run selected; it does not upgrade a Card or held image into stronger+evidence. Prefer the production's supplied or produced pixels, and revise the Run selection when more+representative media becomes available. [Runs](runs.md#use-the-most-representative-visual-evidence-available)+explains the short-lived substitute choices.[Runs and substitutes](runs.md#a-complete-layout-preview) includes a complete Source and Run witha card Candidate, normalization and estimated semantic timing for this kind of layout study.