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(video): align examples with the authoring model
kashorine933cc2e436b
diff --git a/README.md b/README.mdindex fa32b0b2..70127f69 100644--- a/README.md+++ b/README.md@@ -60,6 +60,8 @@ the Skill prepares the executable Hypit Distribution separately; your video proj### UGC+[Generation source](examples/ranking-football/reference.svml) · [Run and production notes](examples/ranking-football/README.md)+<table><tr><th width="16%">Reference</th>@@ -80,6 +82,8 @@ the Skill prepares the executable Hypit Distribution separately; your video proj### Podcast+[Generation source](examples/podcast/reference.svml) · [Run and production notes](examples/podcast/README.md)+<table><tr><th width="16%">Reference</th>@@ -100,6 +104,8 @@ the Skill prepares the executable Hypit Distribution separately; your video proj### Street Interview+[Generation source](examples/interview/reference.svml) · [Run and production notes](examples/interview/README.md)+<table><tr><th width="16%">Reference</th>diff --git a/README.zh-CN.md b/README.zh-CN.mdindex 87b35e3f..c937c5fe 100644--- a/README.zh-CN.md+++ b/README.zh-CN.md@@ -60,6 +60,8 @@ OpenAgents 等 Skill Hub 安装的是同一个 `skills/hypit/` 真目录。首### UGC+[Generation source](examples/ranking-football/reference.svml) · [Run and production notes](examples/ranking-football/README.md)+<table><tr><th width="16%">参考视频</th>@@ -80,6 +82,8 @@ OpenAgents 等 Skill Hub 安装的是同一个 `skills/hypit/` 真目录。首### 播客+[Generation source](examples/podcast/reference.svml) · [Run and production notes](examples/podcast/README.md)+<table><tr><th width="16%">参考视频</th>@@ -100,6 +104,8 @@ OpenAgents 等 Skill Hub 安装的是同一个 `skills/hypit/` 真目录。首### 街头采访+[Generation source](examples/interview/reference.svml) · [Run and production notes](examples/interview/README.md)+<table><tr><th width="16%">参考视频</th>diff --git a/examples/README.md b/examples/README.mdindex 2723ad67..64aacaf4 100644--- a/examples/README.md+++ b/examples/README.md@@ -1,7 +1,24 @@-# Examples+# Video examples-`interview/`, `podcast/` and `ranking-football/` are complete video examples. Each includes SVML,-SVS and SVRun sources, project README guidance and assets; nested directories are independent-variants or explicit reuse projects.+Each of the three formats has one complete generation entry, `reference.svml`, with an adjacent+`reference.svrun`. These Sources generate their main images, recurring voice references and video+Takes, then normalize media, align the Script and compose the finished video. They need no earlier+Build Result or manually supplied presenter, host or scene image.-`minimal-author-package/` is a package-authoring fixture, not a video example.+| Example | What its reference entry demonstrates |+| --- | --- |+| [Football ranking](ranking-football/README.md) | One generated goth host, eight generated comic B-roll images, real player icons, independent speaking Takes and a persistent tier board |+| [Podcast](podcast/README.md) | Related host views, a generated creatine product, holding views, a split opening and a three-scene lifestyle montage |+| [Street interview](interview/README.md) | A shared encounter and two derived close views, motivated cuts, coordinated Moment reveals, and a fresh-footage Caption tracking extension |++Self-contained here means a project directory: SVML, its imported SVS/Kits and the selected local+assets. The installed Hypit Distribution supplies the imported packages. A Runtime Profile and+credentials supply execution; the Sources contain neither. Real identifiers, prepared reveal art,+music and sound effects may remain explicit local assets. Copy the whole example directory when+using it outside this repository.++The other `swap-*` and nested projects preserve separate variations and reuse studies. They have not+all been rebuilt to the same generation standard as these three reference entries. A linked showcase+video demonstrates the original production; a fresh generation is a new result and will differ.++`minimal-author-package/` demonstrates independent component packaging.diff --git a/examples/interview/README.md b/examples/interview/README.mdindex 389cf566..b3ab6974 100644--- a/examples/interview/README.md+++ b/examples/interview/README.md@@ -1,9 +1,99 @@-# Street interview example+# Street interview-This complete example contains the reference project and three independent variants:-`reference.svml` / `reference.svrun`, `swap-host.svml` / `swap-host.svrun`,-`swap-lang.svml` / `swap-lang.svrun`, and `swap-ride.svml` / `swap-ride.svrun`.+[reference.svml](reference.svml) generates the shared encounter and both close views for a black-haired+mob wife beside a sky-blue Lamborghini, interviewed by a young man in a purple shirt and backward+flag cap. [reference.svrun](reference.svrun) targets a finished video without earlier Results.-Read the matching SVML source to see the author graph, the SVS files for recipes and tracking, and-the SVRun file for the build target. The example uses `@hypit/interview-emoji-reveal`, the reusable package in the-repository `packages/` directory.+## From images to the encounter++The shared image establishes both identities, clothing, the car and street. Guest and interviewer+close views each reference that image. The interviewer view keeps a portion of the guest at its left+edge, preserving their spatial relationship. All three are generated; the old `host.png`, `wife.png`+and `guy.png` are not inputs to this entry. Two MiMo Voice Design requests supply recurring voices.++The street-interview Kit consumes interviewer, guest and shared images in that order, followed by+the corresponding interviewer/guest voice references. Three Seedance Mini Takes carry eleven, eight+and eight seconds of dialogue. The first starts with the guest occupied with her bag as the+interviewer approaches. Answers favor her close view; neutral questions can use the shared view,+and surprised questions can cut to him. The final answer ends with her beginning to turn away.+These cuts and actions are authored within the Takes, with no serial tail-frame chain.++Each reveal uses the same Script Moment for the answer strip, sound and colored flash. Prepared+question-mark, sparkles, building and Bitcoin PNGs remain local assets. Their palette works with+Caption and the flashes. The soundtrack and reveal sound are also supplied files. No generic+photographic image prompt stands in for this purpose-made graphic artwork.++## Build a fresh result++[reference.svs](reference.svs) owns this entry's Recipes. The copied Kits are under `kits/`.++```bash+hypit check reference.svrun+hypit measure reference.svml --segment manifest-rule --language en --pace fast --min 4 --max 15 --rounding ceil+hypit measure reference.svml --segment real-estate-rule --language en --pace fast --min 4 --max 15 --rounding ceil+hypit measure reference.svml --segment bitcoin-rule --language en --pace fast --min 4 --max 15 --rounding ceil+hypit plan reference.svrun --runtime ./hypit.runtime.json+```++Those estimates are eleven, eight and eight seconds. Review the selected Endpoints and pricing;+under spending authorization, `hypit build reference.svrun --runtime ./hypit.runtime.json --follow`+generates and composes the work. The initial result uses fixed Caption positions. Watch it to judge+performance, cuts, reading and reveal timing.++## Add head-following Caption to this footage++The published video used externally measured head positions. A fresh generation changes motion,+cuts and actual length, so this Source does not import the original 783-frame `tracking.svs`.+That file describes the old footage only. Supplying it to new Takes would attach plausible-looking+numbers to the wrong video.++After the first Build, export its `final.video` and measure that actual footage, using Google Video+Intelligence with face bounding boxes, another suitable detector, or manual observation. If graphics+obstruct the detector, render a clean picture pass from the same normalized Takes. Use the actual+program frame count and 30 fps clock; identify WIFE across camera cuts, expand face boxes to include+her hair, and write one normalized `[x,y,width,height]` or `null` per frame into a new+`reference-heads.svs`. Keep interpolation within a continuous shot and one identity.++Once that real Recipe exists, add its import to the Source's opening prologue:++```svml+<import as="heads" source="./reference-heads.svs"/>+```++After `vertical` is declared, add its measured timeline and connect it to the existing Caption Track:++```svml+<space:RegionTimeline id="wife-heads" within={vertical} recipe={heads.heads.wife}/>+<caption-fine:Track id="captions" document={story.caption}+ semantic={speech.semantic} program={caption-program} regions={wife-heads}/>+```++Replace the existing `captions` declaration with the connected one. The Recipe contains a WIFE Role+track; `caption.wife` already uses center/bottom anchoring so its text sits above that region.+The untracked BOY continues to use his fixed Style position. A null WIFE region hides her Caption+on that frame; it does not guess a new location.++In the Run, select the first Build's `manifest-rule-semantic.take`, `real-estate-rule-semantic.take`+and `bitcoin-rule-semantic.take` outputs as Candidates. For example, fill the actual Build id in:++```svml+<build-record id="reuse-manifest" build="ACTUAL_BUILD_ID" output="manifest-rule-semantic.take"/>+<satisfy output="manifest-rule-semantic.take" candidate="reuse-manifest"/>+```++Apply the same explicit selection to the other two SemanticTakes, which already contain their media+and alignment. Check `hypit plan` to confirm that the second render contains no new image, voice,+video or alignment requests. Inspect camera cuts and moving head placement in the result. Changing+only Caption or MG needs no repeat generation; changing the footage requires matching measurements.++## Prompt provenance++The author's `studio-prompts-incremental-2026-08-29` export contains the black-haired guest/Lamborghini+shared direction from August 28 at 09:54 and the focused guest/interviewer directions from 13:16–13:21.+The wording and reference counts match the supplied project images and the author's explanation.+This English adaptation keeps the casting, wardrobe and scene, removes an unnecessary phone prop+from the shared prompt, and uses compact gestures in Action. It does not import the earlier blonde+version or the later Porsche adaptation. The voice directions are newly authored; a fresh generation+will differ from the showcased footage.++`swap-host`, `swap-lang` and `swap-ride` are independent variants with their own authored measurements.diff --git a/examples/interview/reference.svml b/examples/interview/reference.svmlindex 7561736b..087e6e0e 100644--- a/examples/interview/reference.svml+++ b/examples/interview/reference.svml@@ -20,8 +20,8 @@<import as="film" from="@hypit/film@1"/><import as="render" from="@hypit/render-hyperframes@1"/><import as="emoji" from="@hypit/interview-emoji-reveal@1"/>- <import as="recipes" source="./recipes.svs"/>- <import as="tracking" source="./tracking.svs"/>+ <import as="recipes" source="./reference.svs"/>+ <import as="image-kit" source="@hypit/gpt-image-kits/phone-ugc-v1"/><import as="interview-kit" source="./kits/street-interview-v1.svs"/><script id="story">@@ -51,36 +51,91 @@</bitcoin-rule></script>- <text:Value id="interviewer-view-prompt">A photograph with the texture of real iPhone footage. Generate a vertical street-interview frame, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. The interviewer is a light-skinned young man with a long side-parted brown hairstyle, a strong nose and jaw, broad shoulders, a purple graphic T-shirt, and a backward American-flag cap. He holds one Hypit microphone toward the guest.</text:Value>- <gpt:Image id="interviewer-view" prompt={interviewer-view-prompt} aspect-ratio="9:16" resolution="2K"/>- <text:Value id="guest-view-prompt">A photograph with the texture of real iPhone footage. Generate a vertical street-interview frame, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. The guest is a light-medium-skinned young woman with a soft oval face, dark almond-shaped eyes, long dark wavy hair, and a defined shoulder line. She wears sunglasses on her head, a black faux-fur jacket over a leopard-print dress, a cross necklace, and carries a structured handbag.</text:Value>- <gpt:Image id="guest-view" prompt={guest-view-prompt} aspect-ratio="9:16" resolution="2K"/>- <text:Value id="shared-view-prompt">A photograph with the texture of real iPhone footage. Generate a vertical street-interview frame, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. Show the light-skinned young interviewer and the light-medium-skinned young woman together beside the blue Lamborghini, preserving their facial features, hair, clothing, microphone, handbag, and left-right relationship.</text:Value>+ <text:Value id="shared-view-shot">+ Generate a close street-interview photograph showing two people from roughly mid-thigh upward.+ The guest stands just left of center, partly facing the camera and looking toward the interviewer on the right.+ Keep their faces, gestures and the handheld microphone readable, with comfortable room around them.+ </text:Value>+ <text:Value id="shared-view-direction">+ On an affluent New York street, a matte sky-blue Lamborghini is parked at the curb with its nose toward+ the left. Its badge is visible low in the frame between the two people standing near the front of the car.+ The guest has a glamorous mob-wife presence. She is about twenty, exceptionally beautiful, with+ Japanese-Italian features, very fair skin and the refined, striking beauty of a Hollywood leading actress.+ Her long black hair falls in loose waves. Sunglasses rest on her head. She wears an expensive-looking dark+ fur coat over a leopard-print strapless short dress, a gold-set cross necklace, and a luxurious structured+ Hermes handbag on her left arm. Her hands rest naturally near her bag, leaving room for an easy gesture.+ The interviewer is a handsome, cheerful young American college student in profile on the right.+ He has tousled curls spilling from a backward American-flag baseball cap, a purple designer graphic T-shirt,+ a studded belt and fashionable flared jeans. He holds one microphone marked HYPIT toward her mouth.+ The stone facades, railings and trees establish a real neighborhood. No other people are in the frame.+ </text:Value>+ <text:Render id="shared-view-prompt" template={image-kit.phone-ugc-v1}>+ <text:Set name="shot" text={shared-view-shot}/>+ <text:Set name="direction" text={shared-view-direction}/>+ </text:Render><gpt:Image id="shared-view" prompt={shared-view-prompt} aspect-ratio="9:16" resolution="2K"/>++ <text:Value id="guest-view-shot">+ Create a closer camera view centered on the woman from the reference. She looks toward the right side of the frame.+ </text:Value>+ <text:Value id="guest-view-direction">+ Preserve her face, black hair, sunglasses, fur coat, leopard-print dress, jewelry and handbag.+ Keep the same street, car and natural light from this new camera position.+ The interviewer's microphone enters from the right toward her mouth; her face and upper body remain clearly visible.+ </text:Value>+ <text:Value id="guest-view-references">+ Use the first image as the complete identity and scene reference. This is another view of the same interview.+ </text:Value>+ <text:Render id="guest-view-prompt" template={image-kit.phone-ugc-v1}>+ <text:Set name="shot" text={guest-view-shot}/>+ <text:Set name="direction" text={guest-view-direction}/>+ <text:Set name="references" text={guest-view-references}/>+ </text:Render>+ <gpt:Image id="guest-view" prompt={guest-view-prompt} aspect-ratio="9:16" resolution="2K">+ <gpt:Reference image={shared-view.image}/>+ </gpt:Image>++ <text:Value id="interviewer-view-shot">+ Create a closer camera view centered on the man from the reference, slightly larger in frame.+ He continues looking left and holding the microphone in the same interviewing pose.+ Keep part of the woman's body visible along the left edge.+ </text:Value>+ <text:Value id="interviewer-view-direction">+ Preserve his handsome face, tousled curls, backward cap, purple shirt and the same street and car.+ Keep the natural light and the left-right relationship between the two people coherent.+ </text:Value>+ <text:Value id="interviewer-view-references">+ Use the first image as the identity and scene reference. Preserve the woman at the edge so this view remains part of their shared space.+ </text:Value>+ <text:Render id="interviewer-view-prompt" template={image-kit.phone-ugc-v1}>+ <text:Set name="shot" text={interviewer-view-shot}/>+ <text:Set name="direction" text={interviewer-view-direction}/>+ <text:Set name="references" text={interviewer-view-references}/>+ </text:Render>+ <gpt:Image id="interviewer-view" prompt={interviewer-view-prompt} aspect-ratio="9:16" resolution="2K">+ <gpt:Reference image={shared-view.image}/>+ </gpt:Image>+<text:Value id="boy-voice-sample">Hey yo! Is this Lamborghini yours?</text:Value><mimo:VoiceDesign id="boy-voice" speech={boy-voice-sample}>A young American man with a bright, casual street-interviewer delivery; energetic, lightly raspy, confident, and conversational, with quick natural pacing.</mimo:VoiceDesign>- <text:Value id="wife-voice-sample">Mmm hmm. Give me three rules to make your first million.</text:Value>+ <text:Value id="wife-voice-sample">That's manifesting, baby.</text:Value><mimo:VoiceDesign id="wife-voice" speech={wife-voice-sample}>A young woman with a playful, self-assured delivery; warm, expressive, slightly husky, and flirtatious, with crisp conversational English.</mimo:VoiceDesign><media:Audio id="soundtrack" src="./assets/shared-soundtrack.m4a"/><media:Audio id="rule-reveal-sound" src="./assets/rule-reveal.wav"/>- <text:Value id="rule-placeholder-prompt">A clean, colorful 3D question-mark icon centered on a simple light background, with crisp edges and no readable text.</text:Value>- <gpt:Image id="rule-placeholder" prompt={rule-placeholder-prompt} aspect-ratio="9:16" resolution="2K"/>- <text:Value id="manifest-icon-prompt">A photograph with the texture of real iPhone footage. Generate a vertical medium close-up, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. A carefully framed photorealistic image with the subject centered and the required object or pose clearly visible. Keep the background concise and unobtrusive.</text:Value>- <gpt:Image id="manifest-icon" prompt={manifest-icon-prompt} aspect-ratio="1:1" resolution="1K"/>- <text:Value id="real-estate-icon-prompt">A photograph with the texture of real iPhone footage. Generate a vertical medium close-up, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. A carefully framed photorealistic image with the subject centered and the required object or pose clearly visible. Keep the background concise and unobtrusive.</text:Value>- <gpt:Image id="real-estate-icon" prompt={real-estate-icon-prompt} aspect-ratio="1:1" resolution="1K"/>- <text:Value id="bitcoin-icon-prompt">A photograph with the texture of real iPhone footage. Generate a vertical medium close-up, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. A carefully framed photorealistic image with the subject centered and the required object or pose clearly visible. Keep the background concise and unobtrusive.</text:Value>- <gpt:Image id="bitcoin-icon" prompt={bitcoin-icon-prompt} aspect-ratio="1:1" resolution="1K"/>+ <!-- Prepared reveal artwork shares the palette of the Caption and flashes. -->+ <media:Image id="rule-placeholder" src="./assets/emoji-reveal/rendered/question-mark.png"/>+ <media:Image id="manifest-icon" src="./assets/emoji-reveal/rendered/sparkles.png"/>+ <media:Image id="real-estate-icon" src="./assets/emoji-reveal/rendered/building-estate.png"/>+ <media:Image id="bitcoin-icon" src="./assets/emoji-reveal/rendered/currency-bitcoin.png"/><text:Value id="manifest-rule-action">BOY is interviewer A, the young man holding the microphone. WIFE is guest B, the woman holding her bag beside the Lamborghini. Preserve their identities, outfits, props, left-right relationship, and the street location throughout.- Throughout the entire take, BOY's long fringe hangs naturally over both eyes and keeps them fully concealed in every camera setup, head turn, gesture, and expression. Both eyes remain invisible behind his hair for the whole clip.- Begin in the exact shared two-person camera setup from @image3 and hold that complete frame for BOY's “Hey yo! Is this Lamborghini yours?” WIFE initially looks down while checking her bag. As BOY asks the question, he takes one step forward toward her. Hearing him, WIFE is briefly startled, then raises her head and looks at BOY.- Hard cut directly to the exact WIFE camera setup from @image2 and hold that complete frame for “Mmm hmm.” WIFE pouts slightly and nods.+ Begin in the exact shared two-person camera setup from @image3 and hold that complete frame for BOY's “Hey yo! Is this Lamborghini yours?” WIFE initially looks down while checking her bag. As BOY asks the question, he takes one step forward toward her. Hearing him, WIFE notices him, raises her head and looks at BOY with casual curiosity.+ Hard cut directly to the exact WIFE camera setup from @image2 and hold that complete frame for “Mmm hmm.” WIFE gives a small knowing smile and nods.Hard cut directly to the exact shared camera setup from @image3 and hold that complete frame for BOY's “Give me three rules to make your first million.” BOY speaks with natural explanatory hand gestures.Hard cut directly to the exact WIFE camera setup from @image2 and hold that complete frame for “OK, the first one is manifest.” WIFE tilts her head and rests one hand lightly against her chin.Hard cut directly to the exact BOY camera setup from @image1 and hold that complete frame for “You manifested a Lamborghini?” BOY makes a visibly surprised hand gesture while asking.@@ -89,7 +144,6 @@<text:Value id="real-estate-rule-action">BOY is interviewer A, the young man holding the microphone. WIFE is guest B, the woman holding her bag beside the Lamborghini. Preserve their identities, outfits, props, left-right relationship, and the street location throughout.- Throughout the entire take, BOY's long fringe hangs naturally over both eyes and keeps them fully concealed in every camera setup, head turn, gesture, and expression. Both eyes remain invisible behind his hair for the whole clip.Begin in the exact shared two-person camera setup from @image3 and hold that complete frame for BOY's “Did it really work?” BOY uses a natural questioning hand gesture.Hard cut directly to the exact WIFE camera setup from @image2 and hold that complete frame for “No. But rule two did.” WIFE shakes her head on “No,” then looks at BOY meaningfully as she delivers the second sentence.Hard cut directly to the exact shared camera setup from @image3 and hold that complete frame for BOY's short line, “Say it.”@@ -100,11 +154,10 @@<text:Value id="bitcoin-rule-action">BOY is interviewer A, the young man holding the microphone. WIFE is guest B, the woman holding her bag beside the Lamborghini. Preserve their identities, outfits, props, left-right relationship, and the street location throughout.- Throughout the entire take, BOY's long fringe hangs naturally over both eyes and keeps them fully concealed in every camera setup, head turn, gesture, and expression. Both eyes remain invisible behind his hair for the whole clip.Begin in the exact shared two-person camera setup from @image3 and hold that complete frame for BOY's “Wait what?” BOY makes a questioning hand gesture.- Hard cut directly to the exact WIFE camera setup from @image2 and hold that complete frame for “Bitcoin. I pressed the wrong button in twenty twelve and made six figures.” WIFE raises one finger while explaining, then makes a broad emphatic hand gesture to show how much money she made.+ Hard cut directly to the exact WIFE camera setup from @image2 and hold that complete frame for “Bitcoin. I pressed the wrong button in twenty twelve and made six figures.” WIFE explains with an easy, open-hand gesture and a small satisfied smile.Hard cut directly to the exact BOY camera setup from @image1 and hold that complete frame for “So you accidentally bought Bitcoin?” BOY reacts with astonished disbelief.- Hard cut directly to the exact WIFE camera setup from @image2 and hold that complete frame for “That's manifesting, baby.” WIFE tilts her head with a cheeky, flirtatious attitude, then turns to her left at the end of the line as if preparing to walk away.+ Hard cut directly to the exact WIFE camera setup from @image2 and hold that complete frame for “That's manifesting, baby.” WIFE gives a cheeky, self-assured smile, then turns to her left at the end of the line as if preparing to walk away.</text:Value><text:Render id="manifest-rule-prompt"@@ -128,7 +181,7 @@<seedance:ReferenceVideo id="manifest-rule-take"model="mini" prompt={manifest-rule-prompt}- duration="10"+ duration="11"resolution="720p" aspect-ratio="9:16" generate-audio="true"><seedance:Reference image={interviewer-view.image}/><seedance:Reference image={guest-view.image}/>@@ -160,15 +213,10 @@</seedance:ReferenceVideo><space:Canvas id="vertical" width="720" height="1280"/>- <space:RegionTimeline id="wife-heads" within={vertical}- recipe={tracking.heads.wife}/><space:Frame id="full-frame" within={vertical}left="0%" top="0%" right="100%" bottom="100%"/><program:Clock id="clock" frame-rate="30"/>- <!-- Preview-only timing policy used by the local mock Run. The paid path- continues to use the WhisperX SemanticTake declarations above. -->-<pipeline:Normalize id="soundtrack-media" source={soundtrack}video="none" audio="default" span-authority="audio" clock={clock}/><pipeline:Normalize id="rule-reveal-sound-media" source={rule-reveal-sound}@@ -219,14 +267,14 @@<caption:Use role="WIFE" style={caption-wife-style}/></caption:Program><caption-fine:Track id="captions" document={story.caption}- semantic={speech.semantic} program={caption-program} regions={wife-heads}/>+ semantic={speech.semantic} program={caption-program}/><emoji:Style id="rule-reveal-style" recipe={recipes.emoji.strip}/><emoji:Track id="rule-reveal" semantic={speech.semantic} canvas={vertical}- style={rule-reveal-style} placeholder={rule-placeholder.image} during="program">- <emoji:Item id="manifest" icon={manifest-icon.image} at={story.moment.manifest}/>- <emoji:Item id="real-estate" icon={real-estate-icon.image} at={story.moment.real-estate}/>- <emoji:Item id="bitcoin" icon={bitcoin-icon.image} at={story.moment.bitcoin}/>+ style={rule-reveal-style} placeholder={rule-placeholder} during="program">+ <emoji:Item id="manifest" icon={manifest-icon} at={story.moment.manifest}/>+ <emoji:Item id="real-estate" icon={real-estate-icon} at={story.moment.real-estate}/>+ <emoji:Item id="bitcoin" icon={bitcoin-icon} at={story.moment.bitcoin}/></emoji:Track><screen:Track id="rule-reveal-flashes" semantic={speech.semantic} canvas={vertical}>diff --git a/examples/interview/reference.svs b/examples/interview/reference.svsnew file mode 100644index 00000000..1d67cc31--- /dev/null+++ b/examples/interview/reference.svs@@ -0,0 +1,134 @@+<?svml using="@hypit/svs@1"?>++<sheet version="1" id="reference">+ interview.street {+ framing: locked-scene;+ pacing: compact;+ performance: natural-street;+ reaction: active;+ gesture: compact;+ }++ speech.visual {+ fit: cover;+ }++ caption.boy {+ stack-order: 70;+ x: 0.5;+ y: 0.63;+ width: 0.88;+ height: 0.22;+ anchor-x: center;+ anchor-y: center;+ align: center;+ block-align: center;+ inline-size: fixed;+ wrap: word;+ max-lines: 2;+ max-words-per-line: 4;+ line-height: 1.08;+ letter-spacing: -0.6;+ word-gap: 12;+ size: 58;+ text-transform: uppercase;+ fill: #FFFFFF;+ stroke-color: #120F0A;+ stroke-width: 3;+ shadow-color: #000000;+ shadow-opacity: 0.95;+ shadow-x: 5;+ shadow-y: 8;+ shadow-blur: 10;+ shadow-spread: 2;+ glow-color: #FFFFFF;+ glow-opacity: 0.94;+ glow-blur: 26;+ glow-spread: 4;+ background: #00000000;+ padding: "0";+ radius: 0;+ karaoke: off;+ cue-enter: spring;+ cue-enter-frames: 4;+ cue-enter-start-scale: 0.75;+ cue-exit: none;+ cue-exit-frames: 0;+ atom-reveal: all;+ lead-frames: 0;+ tail-frames: 3;+ handoff: cut;+ }++ caption.wife {+ stack-order: 70;+ x: 0.5;+ y: 0.63;+ width: 0.88;+ height: 0.22;+ anchor-x: center;+ anchor-y: bottom;+ align: center;+ block-align: center;+ inline-size: fixed;+ wrap: word;+ max-lines: 2;+ max-words-per-line: 4;+ line-height: 1.08;+ letter-spacing: -0.6;+ word-gap: 12;+ size: 58;+ text-transform: uppercase;+ fill: #FFABD4;+ stroke-color: #2A1022;+ stroke-width: 3;+ shadow-color: #000000;+ shadow-opacity: 0.95;+ shadow-x: 5;+ shadow-y: 8;+ shadow-blur: 10;+ shadow-spread: 2;+ glow-color: #FFABD4;+ glow-opacity: 0.94;+ glow-blur: 26;+ glow-spread: 4;+ background: #00000000;+ padding: "0";+ radius: 0;+ karaoke: off;+ cue-enter: spring;+ cue-enter-frames: 4;+ cue-enter-start-scale: 0.75;+ cue-exit: none;+ cue-exit-frames: 0;+ atom-reveal: all;+ lead-frames: 0;+ tail-frames: 3;+ handoff: cut;+ }++ emoji.strip {+ center-x: 0.5;+ top-y: 0.07;+ slot-size: 72;+ slot-gap: 10;+ padding-x: 18;+ padding-y: 14;+ background: #FFFDF7;+ border-color: #161616;+ border-width: 4;+ radius: 22;+ shadow-color: #000000B8;+ shadow-x: 9;+ shadow-y: 10;+ shadow-blur: 0;+ shadow-spread: 0;+ icon-size: 54;+ reveal-frames: 6;+ stack-order: 66;+ }++ film.vertical {+ background: #000000;+ }+</sheet>diff --git a/examples/podcast/README.md b/examples/podcast/README.mdindex 3f526ed0..e729c812 100644--- a/examples/podcast/README.md+++ b/examples/podcast/README.md@@ -1,8 +1,82 @@-# Podcast example+# Creatine podcast-This complete example contains the reference project and three independent variants:-`reference.svml` / `reference.svrun`, `swap-host.svml` / `swap-host.svrun`,-`swap-item.svml` / `swap-item.svrun`, and `swap-app.svml` / `swap-app.svrun`.+[reference.svml](reference.svml) describes the complete image and video generation for a muscular,+beautiful host recommending creatine to a handsome, skinny co-host in glasses. The setting is outside+a Korean restaurant. [reference.svrun](reference.svrun) targets `final.video` without previous Results.-Read each SVML source together with `recipes.svs` and the kit files under `kits/`; each SVRun keeps-the build target separate from authoring and recipe sources.+## The image relationships++```mermaid+flowchart LR+ direction[Main-host direction] --> girl[Main host]+ girl --> guy[Complementary co-host view]+ product[Product direction] --> tub[Creatine tub]+ girl --> girlhold[Main host holding tub]+ tub --> girlhold+ guy --> guyhold[Co-host holding tub]+ tub --> guyhold+ girl --> split[Split-screen opening]+ guy --> split+ girl --> coffee[Cafe selfie]+ girl --> dessert[Dessert-shop mirror selfie]+ girl --> pasta[Courtyard cooking]+```++All nine images are generated. The main host establishes identity and place. The co-host shows the+opposite side of that same conversation, with his own background details and mango bingsu. The two+holding views each use that host's own image plus the same generated tub. The three lifestyle views+branch directly from the main host. None depends on another lifestyle scene or a previous video tail.+The historical `*-hypihub` image files are not inputs to this entry.++The image prompts use the copied phone-image Kit. A product-only prompt directly describes the+creatine packaging. [reference.svs](reference.svs) owns this entry's appearance and video-Kit choices;+`recipes.svs` remains available to the independent variants.++## Performance and coverage++Two MiMo Voice Design requests establish recurring voices. Three Seedance Mini Takes carry the+Script: a five-second split opening, an eight-second product explanation, and a six-second handoff.+The opening uses its own small Text Template in `kits/split-opening.svs`, with dialogue bound from+Script. The other two use the podcast Kit. A silent listener still glances, adjusts posture and reacts.+The tub passes across a cut; the woman is empty-handed on the later return to her view.++One five-second silent B-roll request combines the three lifestyle pictures into a montage. The+second scene is strawberry bingsu, not a literal smoothie demonstration: the montage illustrates the+whole routine. It covers part of the incoming man's response, creating a J-cut and giving the final+cooking scene time to read. Its Recipe uses `once-start`: native-speed playback, truncation when the+Selection is shorter, and an end to coverage when the clip finishes. It does not freeze the last+frame or retime to fill the window. Inspect the newly aligned endpoint and adjust the Selection if+the actual montage needs more reading time. Exact word-to-scene reconstruction would use separate+media Items and adjoining Selections instead.++`shared-soundtrack.m4a` is the only supplied media asset on the final route. Images, voices and Takes+are generated; normalization, alignment, Caption and Film composition are explicit downstream work.++## Run and refine++```bash+hypit check reference.svrun+hypit measure reference.svml --segment opening-question --language en --pace fast --min 4 --max 15 --rounding ceil+hypit measure reference.svml --segment daily-creatine --language en --pace fast --min 4 --max 15 --rounding ceil+hypit measure reference.svml --segment arms-are-asking --language en --pace fast --min 4 --max 15 --rounding ceil+hypit plan reference.svrun --runtime ./hypit.runtime.json+```++The estimates are five, eight and six seconds. Inspect the selected Endpoints and pricing before+paid execution. With spending authorization, use `hypit build reference.svrun --runtime ./hypit.runtime.json --follow`.+Review the actual conversation, product continuity and B-roll handover; reuse produced media for+subsequent Caption or composition changes.++## Prompt provenance++The author's Studio export `studio-prompts-incremental-2026-08-29` contains the main-host prompt+used on August 28 at 09:27, the complementary co-host at 09:40, and the two-host product directions+with two references each. The earlier `studio-prompts-2026-08-28` export contains the creatine+product direction, split layout and cafe/dessert/cooking prompts from August 26–27. Those directions+were translated and adapted into the explicit graph here, preserving grape and mango bingsu in the+host views. The exported records include prompts, parameters and reference counts, but not a complete+mapping from each reference to its selected output; the graph also uses the author's explanation and+the actual project images. The voices are newly authored directions. This is a complete reproducible+request graph, not a claim that a fresh model run will reproduce the published pixels or voices.++`swap-host`, `swap-item` and `swap-app` remain separate variants.diff --git a/examples/podcast/kits/split-opening.svs b/examples/podcast/kits/split-opening.svsnew file mode 100644index 00000000..921fb8e1--- /dev/null+++ b/examples/podcast/kits/split-opening.svs@@ -0,0 +1,17 @@+<?svml using="@hypit/text/svs@1"?>+<sheet version="1" id="split-opening">+ text-template.split-opening { separator: paragraph; }+ text-template.split-opening.block.view {+ kind: fixed; order: 10;+ text: "Animate @image1 as the same fixed horizontal split-screen view, with GUY above and GIRL below. Keep both identities, clothes, microphones, lighting, backgrounds, panel crops and the dividing line. Hold this single layout throughout the opening; no camera movement or internal cuts. GUY uses @audio1 and is the only speaker. GIRL listens silently. Add no captions, UI or floating words.";+ }+ text-template.split-opening.block.dialogue-title {+ kind: fixed; order: 15; text: "Speak the authored dialogue exactly:";+ }+ text-template.split-opening.block.dialogue {+ kind: slot; order: 20; slot: dialogue; optional: false;+ }+ text-template.split-opening.block.action {+ kind: slot; order: 30; slot: action; optional: false;+ }+</sheet>diff --git a/examples/podcast/reference.svml b/examples/podcast/reference.svmlindex bf287f7d..484fe2ff 100644--- a/examples/podcast/reference.svml+++ b/examples/podcast/reference.svml@@ -10,7 +10,6 @@<import as="program" from="@hypit/program-space@1"/><import as="pipeline" from="@hypit/media-pipeline@1"/><import as="whisperx" from="@hypit/whisperx@1"/>- <import as="estimated" from="@hypit/semantic-take-estimate@1"/><import as="speech" from="@hypit/speech-track@1"/><import as="audio-track" from="@hypit/audio-track@1"/><import as="caption" from="@hypit/caption@1"/>@@ -20,7 +19,9 @@<import as="space" from="@hypit/spatial@1"/><import as="film" from="@hypit/film@1"/><import as="render" from="@hypit/render-hyperframes@1"/>- <import as="recipes" source="./recipes.svs"/>+ <import as="recipes" source="./reference.svs"/>+ <import as="image-kit" source="@hypit/gpt-image-kits/phone-ugc-v1"/>+ <import as="opening-kit" source="./kits/split-opening.svs"/><import as="broll-kit" source="./kits/broll-v1.svs"/><import as="podcast-kit" source="./kits/podcast-v1.svs"/>@@ -39,101 +40,249 @@</arms-are-asking></script>- <text:Value id="guy-prompt">A photograph with the texture of real iPhone footage. Generate a vertical medium close-up, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. Young East Asian man with a slim oval face, straight nose, dark almond-shaped eyes behind round black glasses, and thick wavy black hair falling over his forehead. He has a medium shoulder line and wears a pale striped T-shirt with a small black chest mark; keep the microphone and side-facing gaze visible.</text:Value>- <gpt:Image id="guy" prompt={guy-prompt} aspect-ratio="9:16" resolution="2K"/>- <text:Value id="girl-prompt">A photograph with the texture of real iPhone footage. Generate a vertical medium close-up, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. Young East Asian woman with an oval face, dark almond-shaped eyes, straight brows, long straight black hair with wispy bangs, and broad athletic shoulders. She wears a black strapless top and a delicate necklace; keep her upper torso and microphone position visible.</text:Value>+ <!-- The main host establishes the person, place and capture language. -->+ <text:Value id="girl-shot">+ Generate a close half-body podcast view. A woman sits toward the right, her body angled slightly left,+ looking at her conversation partner outside the left edge. Her face remains clearly readable.+ She sits back comfortably in a fabric armchair with her legs crossed, near enough to the camera for her+ face and upper body to feel present, while her shoulders, arms and seated legs stay comfortably framed.+ </text:Value>+ <text:Value id="girl-direction">+ She is exceptionally beautiful, with Japanese-American features and Korean girl-group-idol presence:+ a small refined face, very fair skin, large attractive eyes with blue contact lenses, and long straight+ black hair with blunt bangs. Her strikingly muscular physique contrasts with her delicate face.+ Her shoulders are very broad, her head-to-shoulder proportions excellent, and her arms have pronounced+ muscle definition and visible veins. She wears a black strapless top with an angel-wing design and+ relaxed tweed trousers with an asymmetrical pattern of pale stripes.+ The setting is outside a pleasant Korean street-side restaurant, with textured dark wooden walls,+ plants and other tables. A half-eaten grape bingsu sits on a table in the lower left.+ A podcast microphone extends from the left, marked HYPIT in attractive uppercase lettering.+ Natural light falls comfortably across her face; the wooden wall retains its own color and texture.+ </text:Value>+ <text:Render id="girl-prompt" template={image-kit.phone-ugc-v1}>+ <text:Set name="shot" text={girl-shot}/>+ <text:Set name="direction" text={girl-direction}/>+ </text:Render><gpt:Image id="girl" prompt={girl-prompt} aspect-ratio="9:16" resolution="2K"/>- <text:Value id="guy-hypihub-prompt">A photograph with the texture of real iPhone footage. Generate a vertical medium close-up, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. Young East Asian man with a slim oval face, straight nose, dark almond-shaped eyes behind round black glasses, and thick wavy black hair falling over his forehead. He has a medium shoulder line and wears a pale striped T-shirt with a small black chest mark; keep the microphone and side-facing gaze visible.</text:Value>- <gpt:Image id="guy-hypihub" prompt={guy-hypihub-prompt} aspect-ratio="9:16" resolution="2K"/>- <text:Value id="girl-hypihub-prompt">A photograph with the texture of real iPhone footage. Generate a vertical medium close-up, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. Young East Asian woman with an oval face, dark almond-shaped eyes, straight brows, long straight black hair with wispy bangs, and broad athletic shoulders. She wears a black strapless top and a delicate necklace; keep her upper torso and microphone position visible.</text:Value>- <gpt:Image id="girl-hypihub" prompt={girl-hypihub-prompt} aspect-ratio="9:16" resolution="2K"/>- <text:Value id="split-hypihub-prompt">A photograph with the texture of real iPhone footage. Generate a vertical medium close-up, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. Show the same young East Asian man with thick wavy black hair in the upper half and the same athletic young East Asian woman with long black hair and bangs in the lower half. Keep both faces, shoulders, microphones, and the clean horizontal split visible.</text:Value>- <gpt:Image id="split-hypihub" prompt={split-hypihub-prompt} aspect-ratio="9:16" resolution="2K"/>- <text:Value id="creatine-prompt">A photograph with the texture of real iPhone footage. Generate a vertical medium close-up, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. single creatine supplement tub with a squat cylindrical body, wide lid, and bold product label centered in the frame, with its shape, cap, label placement, and hand scale clearly readable. Use a short neutral product backdrop and no extra objects.</text:Value>- <gpt:Image id="creatine" prompt={creatine-prompt} aspect-ratio="9:16" resolution="2K"/>++ <text:Value id="guy-shot">+ Generate the complementary half-body camera view across this conversation.+ A slim young man sits toward the left, angled and looking right toward the woman outside the frame.+ Match the comfortable camera distance and scale of the first view. The microphone now enters from the right;+ a bowl of mango bingsu on the lower-right table balances him.+ </text:Value>+ <text:Value id="guy-direction">+ He is a handsome, skinny American young man with an endearing nerdy presence, messy curly hair,+ black-framed glasses and an attractive face. He wears a pale finely striped T-shirt with a small abstract+ devil-tail graphic, black trousers and a cool necklace. He sits casually in the corresponding fabric chair.+ Keep the same restaurant, textured wooden surfaces and natural lighting, but show the other part of the+ space with different background objects appropriate to this camera's direction. The microphone reads HYPIT.+ </text:Value>+ <text:Value id="guy-references">+ Use the first image for the restaurant, capture language and conversational geometry. Create the other host and the opposite camera view; do not mirror the woman or duplicate her background.+ </text:Value>+ <text:Render id="guy-prompt" template={image-kit.phone-ugc-v1}>+ <text:Set name="shot" text={guy-shot}/>+ <text:Set name="direction" text={guy-direction}/>+ <text:Set name="references" text={guy-references}/>+ </text:Render>+ <gpt:Image id="guy" prompt={guy-prompt} aspect-ratio="9:16" resolution="2K">+ <gpt:Reference image={girl.image}/>+ </gpt:Image>++ <text:Value id="creatine-prompt">+ A typical e-commerce product photograph of a single creatine supplement tub.+ The main English label reads CREATINE, with a few smaller packaging details arranged like a believable+ supplement label. Show the whole tub and lid clearly against a simple neutral product backdrop.+ Only the tub is present; there are no surrounding objects or floating words.+ </text:Value>+ <gpt:Image id="creatine" prompt={creatine-prompt} aspect-ratio="1:1" resolution="2K"/><text:Value id="guy-voice-sample">OK Sarah, so what is the one single thing that you literally can't live without?</text:Value><mimo:VoiceDesign id="guy-voice" speech={guy-voice-sample}>- A young East Asian man with a curious, friendly podcast-host delivery; clear, relaxed, lightly nasal, and conversational, with an easy medium pace.+ A young American man with a curious, friendly podcast-host delivery; clear, relaxed, lightly nasal, and conversational, with an easy medium pace.</mimo:VoiceDesign><text:Value id="girl-voice-sample">Creatine. Five grams a day, every day.</text:Value><mimo:VoiceDesign id="girl-voice" speech={girl-voice-sample}>- A young East Asian woman with a confident, lively podcast delivery; clear, bright, athletic, and matter-of-fact, with expressive emphasis and natural pacing.+ A young woman speaking conversational American English with a confident, lively podcast delivery; clear, bright, athletic, and matter-of-fact, with expressive emphasis and natural pacing.</mimo:VoiceDesign><asset:Audio id="soundtrack" src="./assets/shared-soundtrack.m4a"/>- <text:Value id="split-prompt">- A photograph with the texture of real iPhone footage. Generate a horizontal split-screen composition, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. Preserve the Rembrandt lighting in the reference images exactly. Create a horizontal split-screen composition with the man in the upper half and the woman in the lower half. Frame both people in tight close-ups focused on their upper bodies from the upper chest upward.+ <text:Value id="split-shot">+ Create a horizontal split-screen image: the man fills the upper panel and the woman fills the lower panel.+ Both panels are close views of their faces and upper bodies, with their microphones visible.</text:Value>-- <text:Value id="guy-hold-prompt">- A photograph with the texture of real iPhone footage. Generate a medium close-up, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. Preserve the Rembrandt lighting from the first reference image exactly. Preserve the microphone and its placement, shifting it only slightly toward the right side of the frame to leave enough room for the man to raise his forearm. Show the man smiling while holding the creatine container from the second reference image. He continues looking toward the right side of the frame as though speaking to someone there. He holds the creatine with the hand on the left side of the frame.- </text:Value>-- <text:Value id="girl-hold-prompt">- A photograph with the texture of real iPhone footage. Generate a medium close-up, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. Preserve the Rembrandt lighting from the first reference image exactly. Preserve the microphone and its placement, shifting it only slightly toward the left side of the frame to leave enough room for the woman to raise her forearm. Show the woman smiling while holding the creatine container from the second reference image. She continues looking toward the left side of the frame as though speaking to someone there. She holds the creatine with the hand on the right side of the frame and points toward it with the hand on the left side of the frame.+ <text:Value id="split-direction">+ Keep each person's identity, styling, restaurant surroundings and natural lighting.+ The man looks right and the woman looks left, as in their individual views.+ Keep the horizontal dividing line clean and preserve the sense that both are present in the conversation.</text:Value>-- <text:Value id="coffee-prompt">- A photograph with the texture of real iPhone footage. Generate an arm's-length selfie, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. Preserve the woman's identity and strongly athletic physique from the reference image, including her visibly muscular arms. She is inside a café, holding the phone high with an extended arm and turning her upper body for a lively selfie. She now wears a light-gray sleeveless hooded zip-up athletic vest over a beige bandeau top, with the hood up and an additional pink baseball cap. She makes an animated, playful expression while pointing behind herself toward a cup of coffee on a table. Other café patrons are subtly visible in the clear background. The lighting is convincingly real: bright afternoon daylight enters from outside and fills the café, casting natural golden side light across her face. Keep the composition energetic and spontaneous, like genuine phone-shot UGC.+ <text:Value id="split-references">+ The first image supplies the man and his camera view for the upper panel. The second supplies the woman and her camera view for the lower panel.</text:Value>+ <text:Render id="split-prompt" template={image-kit.phone-ugc-v1}>+ <text:Set name="shot" text={split-shot}/>+ <text:Set name="direction" text={split-direction}/>+ <text:Set name="references" text={split-references}/>+ </text:Render>+ <gpt:Image id="split" prompt={split-prompt} aspect-ratio="9:16" resolution="2K">+ <gpt:Reference image={guy.image}/>+ <gpt:Reference image={girl.image}/>+ </gpt:Image>- <text:Value id="smoothie-prompt">- A photograph with the texture of real iPhone footage. Generate a mirror selfie, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. Preserve the woman's identity and strongly athletic physique from the reference image, including her visibly muscular arms. She is standing inside a dessert shop and taking a mirror selfie with her phone. She now wears a dark navy Adidas athletic tank top and gray-black over-ear headphones. With her free hand she holds a delicious bowl of strawberry shaved ice close to herself, her lips slightly parted in a lively expression as though she is about to taste it. Her phone has a cute CASETiFY case covered with adorable bunny illustrations, and she looks directly at the phone screen. The lighting is convincingly real: bright afternoon daylight enters from outside and fills the shop, giving her face natural frontal illumination. Use an energetic, playful composition rather than a stiff or symmetrical pose.+ <text:Value id="girl-hold-shot">+ Keep the first image's camera view, framing, natural lighting, seated pose and microphone position.</text:Value>-- <text:Value id="pasta-prompt">- A photograph with the texture of real iPhone footage. Generate a low-angle composition, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. Preserve the woman's identity and strongly athletic physique from the reference image, including her muscular shoulders and arms. She is in the courtyard of her home, photographed from a phone placed on a table and angled upward. The covered courtyard has a warm modern Korean interior style, a wooden slatted roof, and hanging plants. She now wears a cute pink spaghetti-strap apron over a white T-shirt, with part of her hair held back by a white hair clip. She is seriously preparing a batch of macaroni at the stovetop, ready to place it into a pot. The lighting is convincingly real: bright golden afternoon light fills the outdoor space and gives her face natural frontal illumination.+ <text:Value id="girl-hold-direction">+ The woman holds the creatine tub from the second image with the hand on the right side of the frame,+ while the hand on the left points toward it. She still looks left as if talking to her partner.+ Preserve her beauty, identity and muscular physique. The tub belongs naturally in her hand and the scene.</text:Value>-- <text:Value id="lifestyle-montage-story">- Use the three reference images as three ordered shots in one compact, quiet montage. Shot one uses the coffee reference: the woman looks into the camera, raises one relaxed peace sign, and gently turns her face a little from side to side while keeping the rest of her pose natural. Shot two uses the strawberry shaved-ice reference: she continues taking a mirror selfie, keeps the phone clearly visible in one hand and her gaze on its screen, and slightly parts her lips as if she is about to taste the shaved ice. Shot three uses the pasta reference: she calmly continues preparing the pasta with small, practical hand movements and a focused, attentive expression. Keep the action inside each shot continuous and physically coherent, then cut cleanly to the next reference scene in the authored order. Keep all three performances restrained and natural, with ordinary real-time motion, no exaggerated gestures, no repeated blinking, and no slow motion.+ <text:Value id="girl-hold-references">+ Image one supplies the host and the complete scene. Image two supplies the exact product, including its label and proportions.</text:Value>+ <text:Render id="girl-hold-prompt" template={image-kit.phone-ugc-v1}>+ <text:Set name="shot" text={girl-hold-shot}/>+ <text:Set name="direction" text={girl-hold-direction}/>+ <text:Set name="references" text={girl-hold-references}/>+ </text:Render>+ <gpt:Image id="girl-hold" prompt={girl-hold-prompt} aspect-ratio="9:16" resolution="2K">+ <gpt:Reference image={girl.image}/>+ <gpt:Reference image={creatine.image}/>+ </gpt:Image>- <text:Value id="opening-take-prompt">- Preserve the reference image as a fixed horizontal split screen for the entire shot, with the man in the upper half and the woman in the lower half. Keep both identities, outfits, lighting, backgrounds, framing, scale, and the split-screen boundary unchanged. Use one continuous locked view with no cuts, reframing, zoom, or camera movement. In the upper half, the man keeps his gaze toward the right side of the frame throughout while speaking briskly and naturally in one continuous thought, with only a small conversational head movement and no dramatic pause. He says exactly: “OK Sarah so what is the one single thing that you literally can't live without?” At the same time in the lower half, the woman remains completely silent and does not move her mouth. She briefly looks downward, makes one small seated-posture adjustment, then raises her eyes and looks toward the left side of the frame with a focused, attentive expression. Keep both performances understated, simultaneous, photorealistic, and physically coherent. Add no subtitles, captions, labels, UI text, or floating words.+ <text:Value id="guy-hold-shot">+ Keep the first image's camera view, framing, natural lighting, seated pose and microphone position.</text:Value>-- <text:Value id="daily-creatine-action">- ROLE LABEL MAPPING: GIRL is Host A and uses the first image and first audio reference. GUY is Host B and uses the second image and second audio reference. In the dialogue text, every GIRL line belongs only to Host A and every GUY line belongs only to Host B. Only the speaker assigned to the current line moves their mouth; the listener remains silent. GIRL PERFORMANCE: While saying “Creatine,” she lifts the creatine container into a clear, readable position. While saying “Five grams a day, every day,” she points at the container with her free hand and gives small, confident nods. During the next sentence, give each item one distinct, compact gesture with that same free hand: on “I put it in my coffee,” make a small pinching-and-sprinkling motion downward as if adding a scoop to a mug; on “my smoothie,” make one brief circular stirring motion; on “and even my pasta water,” turn the palm outward in a playful, matter-of-fact presentation gesture that lands the joke. Keep the container secure in her other hand throughout. GUY PERFORMANCE: While saying “God, I think you're,” he braces one hand against the sofa and makes one quick seated-posture adjustment, with a mildly incredulous and amused expression. On “treating this as flour,” he raises his right hand for one compact conversational wave. Deliver the word “flour” with a clear falling intonation so the sentence lands decisively rather than sounding like a question. Keep the gestures brisk, readable, understated, and naturally synchronized to the specified phrases.+ <text:Value id="guy-hold-direction">+ The man holds the creatine tub from the second image with the hand on the left side of the frame.+ He still looks right as if speaking to his partner. Keep his other hand naturally at rest.+ Preserve his identity and clothing; the tub belongs naturally in his hand and the scene.</text:Value>-- <text:Value id="arms-are-asking-action">- ROLE LABEL MAPPING: GIRL is Host A and uses the first image and first audio reference. GUY is Host B and uses the second image and second audio reference. In the dialogue text, every GIRL line belongs only to Host A and every GUY line belongs only to Host B. Only the speaker assigned to the current line moves their mouth; the listener remains silent. CONTINUITY: At the beginning, the woman's free hand on the left side of her frame is already lowered naturally while her other hand still holds the creatine container. On the cut to the man, make the container handoff read as one continuous action across the two fixed camera views. Once the man has received it, the container must remain absent from the woman's later shot. PACING AND PERFORMANCE: use quick, conversational handoffs with no long pauses. Keep every expression understated, relaxed, and natural. FIRST GIRL LINE: On “But actually,” she gives a small head tilt and looks toward the left with a slight amused smile, then continues immediately. While saying “you need this way more than I do,” she leans slightly left and extends the container toward the left edge of her frame in one direct offering motion. GUY LINE: On “Wait, what?” he receives the container from the right edge of his frame. On “I didn't ask for that!” he keeps it in one hand, gives one quick palm-up questioning gesture with the free hand, and makes one small confused head shake. FINAL GIRL LINE: The container is no longer visible in her frame. While saying “Of course you did,” she loosely folds her arms and looks left with a slight natural smile. While saying “Look at your arms!” she leans slightly left, unfolds one hand, points toward the left once, and gives one small confident nod. Keep every action brisk, readable, physically coherent, and precisely synchronized to the specified words.+ <text:Value id="guy-hold-references">+ Image one supplies the host and the complete scene. Image two supplies the same product held by the woman.</text:Value>-- <gpt:Image id="split" prompt={split-prompt}- aspect-ratio="9:16" resolution="2K">- <gpt:Reference image={split-hypihub.image}/>- </gpt:Image>-- <gpt:Image id="guy-hold" prompt={guy-hold-prompt}- aspect-ratio="9:16" resolution="2K">+ <text:Render id="guy-hold-prompt" template={image-kit.phone-ugc-v1}>+ <text:Set name="shot" text={guy-hold-shot}/>+ <text:Set name="direction" text={guy-hold-direction}/>+ <text:Set name="references" text={guy-hold-references}/>+ </text:Render>+ <gpt:Image id="guy-hold" prompt={guy-hold-prompt} aspect-ratio="9:16" resolution="2K"><gpt:Reference image={guy.image}/><gpt:Reference image={creatine.image}/></gpt:Image>- <gpt:Image id="girl-hold" prompt={girl-hold-prompt}- aspect-ratio="9:16" resolution="2K">+ <text:Value id="coffee-shot">+ Generate a lively raised-arm selfie of the woman turning toward her phone inside a cafe.+ The view feels comfortably close and casually composed.+ </text:Value>+ <text:Value id="coffee-direction">+ She has changed into a lightweight pale-gray sleeveless zip-front sports layer over a beige bandeau,+ with the hood and an additional pink baseball cap. Her expression is playful and sociable, with one casual+ hand gesture. Preserve her exceptionally attractive face, muscular shoulders and strong arms.+ Other customers are faintly present in the cafe. Afternoon light through the windows gives her face+ natural illumination; cafe furniture and surroundings retain their color.+ </text:Value>+ <text:Value id="coffee-references">+ The first image supplies her identity and physique. Change her clothing, pose and location to the cafe selfie described here.+ </text:Value>+ <text:Render id="coffee-prompt" template={image-kit.phone-ugc-v1}>+ <text:Set name="shot" text={coffee-shot}/>+ <text:Set name="direction" text={coffee-direction}/>+ <text:Set name="references" text={coffee-references}/>+ </text:Render>+ <gpt:Image id="coffee" prompt={coffee-prompt} aspect-ratio="9:16" resolution="2K"><gpt:Reference image={girl.image}/>- <gpt:Reference image={creatine.image}/>- </gpt:Image>-- <gpt:Image id="coffee" prompt={coffee-prompt}- aspect-ratio="9:16" resolution="2K">- <gpt:Reference image={girl-hypihub.image}/></gpt:Image>- <gpt:Image id="smoothie" prompt={smoothie-prompt}- aspect-ratio="9:16" resolution="2K">+ <text:Value id="smoothie-shot">+ Generate a lively mirror selfie of the woman standing in a dessert shop.+ She holds her phone in one hand, watches its screen, and holds a bowl of strawberry bingsu in the other.+ </text:Value>+ <text:Value id="smoothie-direction">+ She wears a navy Adidas sports tank and charcoal over-ear headphones. Her phone has a playful CASETiFY+ case with a cute rabbit design. She parts her lips with an amused expression as if about to taste the bingsu.+ Keep her very beautiful face and clearly muscular arms and shoulders. The shop is lit naturally through+ its windows, with a readable background. The composition feels spontaneous rather than rigid.+ </text:Value>+ <text:Value id="smoothie-references">+ Preserve the woman from the first image while creating this different outfit, activity and location. This dessert scene broadly supports her food routine; it is not a literal smoothie demonstration.+ </text:Value>+ <text:Render id="smoothie-prompt" template={image-kit.phone-ugc-v1}>+ <text:Set name="shot" text={smoothie-shot}/>+ <text:Set name="direction" text={smoothie-direction}/>+ <text:Set name="references" text={smoothie-references}/>+ </text:Render>+ <gpt:Image id="smoothie" prompt={smoothie-prompt} aspect-ratio="9:16" resolution="2K"><gpt:Reference image={girl.image}/></gpt:Image>- <gpt:Image id="pasta" prompt={pasta-prompt}- aspect-ratio="9:16" resolution="2K">+ <text:Value id="pasta-shot">+ Generate a low-angle photograph as if a phone has been set on a table in the woman's home courtyard.+ She is close enough for her attentive expression and practical hand movements to be clearly readable.+ </text:Value>+ <text:Value id="pasta-direction">+ She wears a pink spaghetti-strap apron over a textured pale T-shirt, with a pale hair clip holding back+ part of her hair. She is preparing macaroni at a stovetop, with pasta ready to go into a pot.+ Keep her exceptionally attractive face, strong arms and broad muscular shoulders.+ The covered courtyard has a warm modern Korean domestic feel, with a slatted wooden roof and hanging plants.+ Afternoon sunlight enters the courtyard and gives her face natural illumination, while the covered area+ and surrounding wood keep the scene tonally grounded.+ </text:Value>+ <text:Value id="pasta-references">+ Use the first image for her identity and physique; create the home-cooking scene independently of the other lifestyle images.+ </text:Value>+ <text:Render id="pasta-prompt" template={image-kit.phone-ugc-v1}>+ <text:Set name="shot" text={pasta-shot}/>+ <text:Set name="direction" text={pasta-direction}/>+ <text:Set name="references" text={pasta-references}/>+ </text:Render>+ <gpt:Image id="pasta" prompt={pasta-prompt} aspect-ratio="9:16" resolution="2K"><gpt:Reference image={girl.image}/></gpt:Image>+ <text:Value id="lifestyle-montage-story">+ Use @image1, @image2 and @image3 as three ordered shots in one five-second montage about her daily food routine.+ Give all three scenes readable screen time, including the final cooking scene.+ In the cafe selfie she gives a small playful head turn. In the mirror selfie she keeps her phone in one hand,+ looks at its screen and brings the bingsu slightly closer with an amused expression.+ In the courtyard she continues preparing pasta with small, practical movements.+ Cut cleanly between the three scenes; each scene is continuous internally. Keep her identity, outfit and+ setting faithful to that scene's reference. Motion is ordinary real-time motion, relaxed and physically coherent.+ </text:Value>++ <text:Value id="opening-action">+ The man speaks briskly and naturally while looking right, with a small conversational head movement.+ At the same time the woman remains silent, briefly glances down, makes one small seated-posture adjustment,+ and looks back left attentively. Her lips do not form his words. Keep both performances understated and alive.+ </text:Value>+ <text:Render id="opening-take-prompt" template={opening-kit.split-opening}>+ <text:Set name="dialogue" text={story.segment.opening-question.dialogue}/>+ <text:Set name="action" text={opening-action}/>+ </text:Render>++ <text:Value id="daily-creatine-action">+ GIRL is Host A (@image1, @audio1); GUY is Host B (@image2, @audio2).+ Only the current speaker speaks; the listener reacts silently.+ GIRL presents the creatine tub with cheerful, matter-of-fact confidence. On "Creatine", make its label easy+ to see; on the daily amount, give it one light indicating gesture and a small nod. Keep it secure in that hand.+ Her list of foods has a playful, almost obvious-to-her rhythm, with a relaxed illustrative gesture from the free hand.+ GUY shifts his seated posture with mild amused disbelief, then makes one compact open-hand gesture as+ "treating this as flour" lands. Let "flour" finish with falling intonation. Keep the banter brisk and natural;+ no finger counting or exaggerated reactions.+ </text:Value>++ <text:Value id="arms-are-asking-action">+ GIRL is Host A (@image1, @audio1); GUY is Host B (@image2, @audio2).+ Keep the exchange quick, relaxed and gently teasing. Each person speaks only their own lines.+ At first GIRL holds the tub and looks left with an amused half-smile. As she says "you need this way more+ than I do", she offers it toward the left edge of her frame. Cut on that offering action to GUY's view,+ where he receives it from the right; his holding reference supplies the resulting pose.+ GUY gives a small confused head shake and one palm-up gesture with his free hand while keeping the tub secure.+ On the return to GIRL, her hands are empty. She loosely folds her arms, then gestures toward him once as+ "Look at your arms!" lands with a confident nod. Make the change of possession clear through the cuts.+ </text:Value>+<text:Render id="lifestyle-montage-prompt"template={broll-kit.broll-v1} recipe={recipes.broll.calm-ugc-montage}><text:Set name="story" text={lifestyle-montage-story}/>@@ -160,7 +309,7 @@</text:Render><seedance:ReferenceVideo id="opening-take" model="mini"- prompt={opening-take-prompt} duration="4"+ prompt={opening-take-prompt} duration="5"resolution="720p" aspect-ratio="9:16" generate-audio="true"><seedance:Reference image={split.image}/><seedance:Reference audio={guy-voice.reference}/>@@ -208,15 +357,6 @@<whisperx:SemanticTake id="arms-are-asking-semantic" narrative={story}segment={story.segment.arms-are-asking} media={arms-are-asking-media.media} language="en"/>- <estimated:SemanticTake id="opening-estimated" narrative={story}- segment={story.segment.opening-question} media={opening-media.media}- policy={recipes.speech.preview}/>- <estimated:SemanticTake id="daily-creatine-estimated" narrative={story}- segment={story.segment.daily-creatine} media={daily-creatine-media.media}- policy={recipes.speech.preview}/>- <estimated:SemanticTake id="arms-are-asking-estimated" narrative={story}- segment={story.segment.arms-are-asking} media={arms-are-asking-media.media}- policy={recipes.speech.preview}/><speech:Track id="speech" visual-frame={full-frame}visual-appearance={recipes.speech.visual} visual-z="10">diff --git a/examples/podcast/reference.svs b/examples/podcast/reference.svsnew file mode 100644index 00000000..4a282d9d--- /dev/null+++ b/examples/podcast/reference.svs@@ -0,0 +1,152 @@+<?svml using="@hypit/svs@1"?>++<sheet version="1" id="reference">+ broll.calm-ugc-montage {+ material-mode: practical-real;+ story-shape: montage-sequence;+ edit-language: jump-cut-beats;+ camera-language: reference-locked;+ motion-intensity: readable;+ }++ podcast.daily-creatine {+ framing: locked-view;+ edit-language: speaker-cuts;+ pacing: compact;+ performance: natural-banter;+ reaction: subtle-listening;+ gesture: natural;+ }++ podcast.arms-are-asking {+ framing: locked-view;+ edit-language: speaker-cuts;+ pacing: compact;+ performance: natural-banter;+ reaction: subtle-listening;+ gesture: natural;+ }++ speech.visual {+ fit: cover;+ }++ media.lifestyle {+ stack-order: 40;+ fit: cover;+ playback: once-start;+ clip: frame;+ }++ caption.original-girl {+ stack-order: 70;+ x: 0.5;+ y: 0.5;+ width: 0.88;+ height: 0.18;+ anchor-x: center;+ anchor-y: center;+ align: center;+ block-align: center;+ inline-size: fixed;+ wrap: word;+ max-lines: 2;+ max-words-per-line: 4;+ line-height: 1.16;+ letter-spacing: -0.8;+ word-gap: 12;+ size: 52;+ text-transform: uppercase;+ fill: #FFFFFF;+ stroke-color: #170A20;+ stroke-width: 3;+ shadow-color: #000000;+ shadow-opacity: 0.82;+ shadow-x: 4;+ shadow-y: 5;+ shadow-blur: 4;+ shadow-spread: 1;+ glow-color: #F16BFF;+ glow-opacity: 0.42;+ glow-blur: 10;+ glow-spread: 1;+ background: #00000000;+ padding: "0";+ radius: 0;+ karaoke: current;+ karaoke-transition: step;+ active-fill: #F16BFF;+ active-stroke-color: #170A20;+ active-stroke-width: 3;+ active-glow-color: #F16BFF;+ active-glow-opacity: 0.96;+ active-glow-blur: 14;+ active-glow-spread: 2;+ active-response: none;+ cue-enter: pop;+ cue-enter-frames: 4;+ cue-exit: fade;+ cue-exit-frames: 2;+ lead-frames: 2;+ tail-frames: 10;+ handoff: cut;+ }++ caption.original-guy {+ stack-order: 70;+ x: 0.5;+ y: 0.5;+ width: 0.88;+ height: 0.18;+ anchor-x: center;+ anchor-y: center;+ align: center;+ block-align: center;+ inline-size: fixed;+ wrap: word;+ max-lines: 2;+ max-words-per-line: 4;+ line-height: 1.16;+ letter-spacing: -0.8;+ word-gap: 12;+ size: 52;+ text-transform: uppercase;+ fill: #FFFFFF;+ stroke-color: #170A20;+ stroke-width: 3;+ shadow-color: #000000;+ shadow-opacity: 0.82;+ shadow-x: 4;+ shadow-y: 5;+ shadow-blur: 4;+ shadow-spread: 1;+ glow-color: #77D9FF;+ glow-opacity: 0.42;+ glow-blur: 10;+ glow-spread: 1;+ background: #00000000;+ padding: "0";+ radius: 0;+ karaoke: current;+ karaoke-transition: step;+ active-fill: #77D9FF;+ active-stroke-color: #170A20;+ active-stroke-width: 3;+ active-glow-color: #77D9FF;+ active-glow-opacity: 0.96;+ active-glow-blur: 14;+ active-glow-spread: 2;+ active-response: none;+ cue-enter: pop;+ cue-enter-frames: 4;+ cue-exit: fade;+ cue-exit-frames: 2;+ lead-frames: 2;+ tail-frames: 10;+ handoff: cut;+ }++ film.vertical {+ background: #000000;+ }+</sheet>diff --git a/examples/ranking-football/README.md b/examples/ranking-football/README.mdindex 2baa2cc6..a2568d5c 100644--- a/examples/ranking-football/README.md+++ b/examples/ranking-football/README.md@@ -1,9 +1,55 @@-# Football ranking example+# Football ranking-This complete example contains the reference project, three direct variants, and the additional-banana/reference-sync variants under their own directories. Each project directory includes its-SVML author source, SVS recipes/kits, and SVRun build or reuse sources.+[reference.svml](reference.svml) generates a goth football commentator and all eight comedy B-roll+images, then makes two independent speaking Takes from the same presenter image and designed voice.+A tier board preserves its state while the argument, Caption and overlapping image inserts unfold.+[reference.svrun](reference.svrun) targets `final.video` without earlier Results or replacement files.-Read the top-level `reference.svml`, `swap-host.svml`, `swap-effect.svml`, and `swap-topic.svml` with-their matching `.svrun` files first. The nested banana projects show explicit reuse and replacement-of generated media.+## What is generated and what is supplied++- `presenter.image`: the striking goth host, seated to the right in a sky-blue football classroom.+- Eight `broll-*.image` outputs: independent text-to-image jokes for the annotated Script passages.+- `presenter-voice.reference`: one MiMo Voice Design request reused by both Takes.+- `ronaldo-take.video` and `messi-take.video`: Seedance Mini performances with audio, followed by+ normalization and WhisperX alignment. Both use the same presenter image; there is no tail-frame chain.+- Eleven supplied player photographs identify the board entries. Nine are presets and two enter+ during their spoken Segments. These real identifiers are intentionally not generated lookalikes.+- `shared-soundtrack.m4a`, `ranking-appear.wav` and `ranking-move.wav` are supplied audio assets.++The image prompts, reference bindings and Action are in the Source. The copied image/speaker Kits+are in `kits/`; [recipes.svs](recipes.svs) owns appearance and the speaker's pause-trim direction.+The speaker Kit requests natural edited delivery; it does not perform a postprocessing trim.+The B-roll uses its own visual language: theatrical tragedy, product comedy, sports photography or+parody UI. It does not automatically inherit the host's iPhone capture paragraph.++## Work with the example++With a configured Runtime and installed Distribution, run from this directory:++```bash+hypit check reference.svrun+hypit measure reference.svml --segment ronaldo --language en --pace fast --min 4 --max 15 --rounding ceil+hypit measure reference.svml --segment messi --language en --pace fast --min 4 --max 15 --rounding ceil+hypit plan reference.svrun --runtime ./hypit.runtime.json+```++Both passages estimate to nine seconds at that delivery policy; the authored ten-second requests+allow room for expression. Read the plan and its Endpoint/pricing information before authorizing paid+work. Under that authorization, `hypit build reference.svrun --runtime ./hypit.runtime.json --follow`+produces the final video. Watch the actual result and reuse its media outputs when refining graphics.++When changing the ranked subjects, use supplied photographs or search for recognizable real+photographs, official logos or product assets. Record source pages and any relevant attribution with+the project, then crop or pad consistently for the board. The original project's player files did+not include a source URL log; this example does not invent one. No asset lookup occurs during Build.++## Prompt provenance and scope++The presenter direction is an English adaptation of the author's supplied Messi goth-girl prompt,+including broad shoulders, the HYPIT spiked cap, right-side placement, a balancing desk and abstract+football posters. Output shape stays in the image parameters. The eight comic image prompts were+retained from the original ranking Source. Action corrects the host's pronouns and uses a few+meaningful gestures while preserving the dry, affectionate football humor.++`swap-host`, `swap-effect`, `swap-topic` and the nested banana projects are separate studies. Their+Sources and Runs are not inputs to this complete generation entry.diff --git a/examples/ranking-football/reference.svml b/examples/ranking-football/reference.svmlindex b0940fe2..0c3f1975 100644--- a/examples/ranking-football/reference.svml+++ b/examples/ranking-football/reference.svml@@ -8,7 +8,6 @@<import as="gpt" from="@hypit/gpt-image@1"/><import as="seedance" from="@hypit/seedance@1"/><import as="pipeline" from="@hypit/media-pipeline@1"/>- <import as="estimated" from="@hypit/semantic-take-estimate@1"/><import as="whisperx" from="@hypit/whisperx@1"/><import as="speech" from="@hypit/speech-track@1"/><import as="media-track" from="@hypit/media-track@1"/>@@ -22,6 +21,7 @@<import as="film" from="@hypit/film@1"/><import as="render" from="@hypit/render-hyperframes@1"/><import as="recipes" source="./recipes.svs"/>+ <import as="image-kit" source="@hypit/gpt-image-kits/phone-ugc-v1"/><import as="speaker-kit" source="./kits/speaker-v1.svs"/><!-- The Program projection owns the complete board lifetime. Each discussed@@ -36,8 +36,31 @@</messi></script>- <text:Value id="presenter-prompt">A photograph with the texture of real iPhone footage. Generate a vertical medium close-up, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. A pale young woman with a narrow oval face, large gray eyes, straight brows, long jet-black hair with blunt bangs, and a slim shoulder line wears a black spiked cap, black choker, layered silver necklaces, and a black graphic top while holding a furry microphone. Keep her face, hair, shoulders, and microphone visible.</text:Value>+ <text:Value id="presenter-shot">+ Generate a close half-body view of a woman seated to the right, speaking to the camera with a handheld microphone.+ Keep her head, broad shoulders and arms comfortably inside the frame, with her seated legs visible below her torso.+ She is comfortably near the camera, with natural head-to-body proportions. Her face sits toward the upper right.+ A small desk at her left balances the composition and leaves useful space for the player's icon on the left.+ </text:Value>+ <text:Value id="presenter-direction">+ She is exceptionally beautiful, with Japanese-American features and the captivating presence of a Korean girl-group idol.+ She has a small, refined face, very fair skin, broad shoulders and excellent head-to-shoulder proportions.+ Her light goth makeup has a sweet, cool attitude and a feline quality around the eyes.+ Her long black hair is straight with blunt bangs. She wears attractive earrings, a choker and layered jewelry,+ a spiked baseball cap marked HYPIT in uppercase Gothic lettering, and a black cotton goth-style short-sleeved top.+ She sits on a chair in a university classroom. The desk beside her has a cluttered open storage compartment.+ Behind her are a classroom door and casually arranged football posters in varied sizes:+ a sky-blue-and-white striped jersey, a football, a Ballon d'Or and an abstract football design.+ There are no players, school names or school logos on the posters.+ Sky-blue stripes, paper and textured pale wall surfaces establish the classroom palette;+ her black outfit and hair give her a clear presence against it.+ </text:Value>+ <text:Render id="presenter-prompt" template={image-kit.phone-ugc-v1}>+ <text:Set name="shot" text={presenter-shot}/>+ <text:Set name="direction" text={presenter-direction}/>+ </text:Render><gpt:Image id="presenter" prompt={presenter-prompt} aspect-ratio="9:16" resolution="2K"/>+<text:Value id="presenter-voice-sample">Welcome back. Today we're ranking the most ridiculous football takes.</text:Value><mimo:VoiceDesign id="presenter-voice" speech={presenter-voice-sample}>A young woman with a smoky, low-register alt voice and dry sarcastic presenter delivery; cool, composed, slightly deadpan, with sharp comedic timing and confident emphasis.@@ -45,64 +68,62 @@<asset:Audio id="soundtrack" src="./assets/shared-soundtrack.m4a"/><asset:Audio id="ranking-appear-sfx" src="./assets/ranking-appear.wav"/><asset:Audio id="ranking-move-sfx" src="./assets/ranking-move.wav"/>- <text:Value id="ronaldo-icon-prompt">A photograph with the texture of real iPhone footage. Generate a close portrait, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. tan Portuguese footballer in his late thirties with an angular rectangular face, high cheekbones, thick dark eyebrows, narrow brown eyes, a straight prominent nose, a square jaw, and a clean-shaven face. His short dark hair is faded at the sides and brushed sharply back from the forehead; he has a muscular neck, very broad athletic shoulders, and a focused expression. He wears a plain red shirt.. Keep the portrait centered, shoulders visible, and use a simple neutral background with no readable text or logos.</text:Value>- <gpt:Image id="ronaldo-icon" prompt={ronaldo-icon-prompt} aspect-ratio="1:1" resolution="1K"/>- <text:Value id="messi-icon-prompt">A photograph with the texture of real iPhone footage. Generate a close portrait, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. light-skinned Argentine footballer in his late thirties with a rounded oval face, soft brown eyes, thick brows, a straight nose, and a full neatly trimmed brown beard. His short brown hair is cropped close at the sides and brushed forward on top; he has a compact athletic neck and shoulders and a calm expression. He wears a plain sky-blue shirt.. Keep the portrait centered, shoulders visible, and use a simple neutral background with no readable text or logos.</text:Value>- <gpt:Image id="messi-icon" prompt={messi-icon-prompt} aspect-ratio="1:1" resolution="1K"/>- <text:Value id="yamal-icon-prompt">A photograph with the texture of real iPhone footage. Generate a close portrait, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. young Spanish footballer with medium-brown skin, a youthful round face, dark almond-shaped eyes, thick brows, a broad nose, full lips, and a bright open expression. His short tight dark curls have a clean fade at the sides; he has a lean athletic neck and shoulders. He wears a plain blue-and-red shirt.. Keep the portrait centered, shoulders visible, and use a simple neutral background with no readable text or logos.</text:Value>- <gpt:Image id="yamal-icon" prompt={yamal-icon-prompt} aspect-ratio="1:1" resolution="1K"/>- <text:Value id="haaland-icon-prompt">A photograph with the texture of real iPhone footage. Generate a close portrait, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. very tall fair-skinned Scandinavian footballer with a long rectangular face, pale blue eyes, straight light brows, a broad forehead, a straight nose, and a clean-shaven jaw. His long blond hair is parted near the center and falls past his ears; he has an unusually long neck and very broad shoulders and a neutral expression. He wears a plain light-blue shirt.. Keep the portrait centered, shoulders visible, and use a simple neutral background with no readable text or logos.</text:Value>- <gpt:Image id="haaland-icon" prompt={haaland-icon-prompt} aspect-ratio="1:1" resolution="1K"/>- <text:Value id="de-bruyne-icon-prompt">A photograph with the texture of real iPhone footage. Generate a close portrait, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. fair-skinned Belgian footballer in his early thirties with an angular long face, pale blue eyes, light brows, a straight nose, and light ginger stubble along the jaw. His short ginger-blond hair is brushed back with slightly tousled texture; he has a lean athletic neck and shoulders and an intent expression. He wears a plain sky-blue shirt.. Keep the portrait centered, shoulders visible, and use a simple neutral background with no readable text or logos.</text:Value>- <gpt:Image id="de-bruyne-icon" prompt={de-bruyne-icon-prompt} aspect-ratio="1:1" resolution="1K"/>- <text:Value id="bellingham-icon-prompt">A photograph with the texture of real iPhone footage. Generate a close portrait, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. young Black English footballer with medium-dark brown skin, an oval face, dark brown eyes, thick brows, a broad nose, full lips, and a broad friendly smile. His short tight curls are shaped into a neat taper fade; he has a strong athletic neck and broad shoulders. He wears a plain white shirt.. Keep the portrait centered, shoulders visible, and use a simple neutral background with no readable text or logos.</text:Value>- <gpt:Image id="bellingham-icon" prompt={bellingham-icon-prompt} aspect-ratio="1:1" resolution="1K"/>- <text:Value id="kane-icon-prompt">A photograph with the texture of real iPhone footage. Generate a close portrait, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. fair-skinned English footballer in his early thirties with a broad square face, blue-gray eyes, thick brows, a straight nose, and short ginger-brown stubble. His short brown hair is brushed forward with a subtle side part; he has a strong neck, broad athletic shoulders, and a serious expression. He wears a plain white shirt.. Keep the portrait centered, shoulders visible, and use a simple neutral background with no readable text or logos.</text:Value>- <gpt:Image id="kane-icon" prompt={kane-icon-prompt} aspect-ratio="1:1" resolution="1K"/>- <text:Value id="modric-icon-prompt">A photograph with the texture of real iPhone footage. Generate a close portrait, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. light-skinned Croatian footballer in his late thirties with a narrow elongated face, light blue eyes, fine brows, a straight nose, and light stubble. His long light-brown hair is parted in the middle and falls to his shoulders; he has a lean neck and slim athletic shoulders and a composed expression. He wears a plain white shirt.. Keep the portrait centered, shoulders visible, and use a simple neutral background with no readable text or logos.</text:Value>- <gpt:Image id="modric-icon" prompt={modric-icon-prompt} aspect-ratio="1:1" resolution="1K"/>- <text:Value id="mbappe-icon-prompt">A photograph with the texture of real iPhone footage. Generate a close portrait, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. young Black French footballer with deep brown skin, a broad oval face, dark brown eyes, thick brows, a broad nose, a strong jaw, and full lips. His black hair is close-cropped with a clean fade; he has a muscular neck, powerful athletic shoulders, and a focused expression. He wears a plain navy shirt.. Keep the portrait centered, shoulders visible, and use a simple neutral background with no readable text or logos.</text:Value>- <gpt:Image id="mbappe-icon" prompt={mbappe-icon-prompt} aspect-ratio="1:1" resolution="1K"/>- <text:Value id="neymar-icon-prompt">A photograph with the texture of real iPhone footage. Generate a close portrait, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. tan Brazilian footballer in his early thirties with an oval face, expressive brown eyes, thick brows, a straight nose, defined cheekbones, and a neatly trimmed beard and mustache. His short dark hair has faded sides and a textured top swept slightly back; he has an athletic neck and shoulders and a lively expression. He wears a plain yellow shirt.. Keep the portrait centered, shoulders visible, and use a simple neutral background with no readable text or logos.</text:Value>- <gpt:Image id="neymar-icon" prompt={neymar-icon-prompt} aspect-ratio="1:1" resolution="1K"/>- <text:Value id="benzema-icon-prompt">A photograph with the texture of real iPhone footage. Generate a close portrait, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. olive-skinned French-Algerian footballer in his mid-thirties with a broad square face, dark brown eyes, thick brows, a strong nose, heavy jaw, and a full dense beard. His head is shaved close; he has a thick neck, very broad muscular shoulders, and a calm serious expression. He wears a plain white shirt.. Keep the portrait centered, shoulders visible, and use a simple neutral background with no readable text or logos.</text:Value>- <gpt:Image id="benzema-icon" prompt={benzema-icon-prompt} aspect-ratio="1:1" resolution="1K"/>-- <!-- Real Seedance Takes use explicit fixed ten-second requests. These- estimates belong only to the optional static preview candidates. -->+ <!-- Existing player photographs identify the board entries. Joke images below are generated. -->+ <asset:Image id="ronaldo-icon" src="./assets/ronaldo.jpeg"/>+ <asset:Image id="messi-icon" src="./assets/messi.jpeg"/>+ <asset:Image id="yamal-icon" src="./assets/yamal.jpg"/>+ <asset:Image id="haaland-icon" src="./assets/haaland.jpg"/>+ <asset:Image id="de-bruyne-icon" src="./assets/debrunye.jpg"/>+ <asset:Image id="bellingham-icon" src="./assets/bellingham-new.jpg"/>+ <asset:Image id="kane-icon" src="./assets/kane.jpg"/>+ <asset:Image id="modric-icon" src="./assets/modric.jpg"/>+ <asset:Image id="mbappe-icon" src="./assets/mbappe-new.jpg"/>+ <asset:Image id="neymar-icon" src="./assets/neymar.jpg"/>+ <asset:Image id="benzema-icon" src="./assets/benzema.jpg"/><text:Value id="ronaldo-action">- Keep the handheld microphone in the same hand throughout, and use only the free hand for gestures. At the opening, widen his eyes and nod emphatically as he begins speaking. On "hair gel", deliberately run the free hand over his hair. On "Greek god", react with exaggerated surprise. On "Greek tragedy", shake his head with a mocking laugh. On "But hey", lean his body forward and raise the free index finger as if he has just thought of something. On the final sentence, give a sarcastic shrug.+ Keep the microphone in the same hand and gesture with the free hand. She delivers the ranking with dry,+ self-assured amusement and brisk conversational timing. Begin with one confident nod.+ On "hair gel", briefly smooth her own hair as the visual joke lands. "Greek god" earns an appraising look;+ "Greek tragedy" turns it into a small disbelieving head shake and half-smile.+ On "But hey", lean in slightly as if offering one last concession, then finish with a sarcastic shrug.+ Let the expression change with the thought; gestures need not fill every word.</text:Value><text:Value id="messi-action">- Keep the handheld microphone in the same hand throughout, and use only the free hand for gestures. Throughout the opening sentence, nod continuously in small rhythmic beats. On "tired dad", look both admiring and incredulous. On "Walks around", draw a small circle in the air with the free hand. On "touches the ball", point the free thumb toward the right side of frame. As he says the exact phrase "hat trick", look deeply impressed and lift and extend the free arm upward above shoulder level with the palm open toward the upper part of the frame. The prosody of "hat trick" is strict: put the primary lexical stress on HAT. Make HAT clearly higher, louder, longer, and more accented; make "trick" shorter, lighter, and lower. The pitch must fall sharply from HAT through "trick" and finish low and closed, as a confident statement. Do not stress TRICK, do not keep both words level, and absolutely do not use rising or question-like intonation. On the final sentence, give one firm affirmative nod.+ Keep the microphone in the same hand and use only the free hand for gestures.+ She sounds genuinely impressed while keeping the dry humor of the first passage.+ On "tired dad", give a small affectionate smile of disbelief. Let a relaxed outward gesture accompany+ "Walks around", then become more animated as the surprising result arrives.+ On "hat trick", lift the free hand in one open, impressed gesture. Stress HAT and let "trick" fall naturally+ so the phrase lands as a confident statement. Finish "Definitely S tier" with one decisive nod.+ Do not mime exact numbers with fingers.</text:Value><!-- Comedy B-roll source images. Aspect ratio and resolution stay modelparameters; semantic placement is authored separately in the Script. --><text:Value id="broll-hair-gel-prompt">- A photograph with the texture of real iPhone footage. Generate a carefully framed composition, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. An absurdly polished commercial product photograph of a fictional men's hair gel: an unbranded black jar with a glossy ribbon of translucent gel squeezed out beside it. Next to the jar is the back of a man's head, cropped at the shoulders, with a sharply styled short haircut so overloaded with gel that every strand looks wet, mirror-shiny, and almost reflective. Crisp studio lighting, premium cosmetics advertising, realistic textures, deadpan visual comedy, no readable text, no logos.+ An absurdly polished commercial product photograph of a fictional men's hair gel: an unbranded black jar with a glossy ribbon of translucent gel squeezed out beside it. Next to the jar is the back of a man's head, cropped at the shoulders, with a sharply styled short haircut so overloaded with gel that every strand looks wet, mirror-shiny, and almost reflective. Crisp studio lighting, premium cosmetics advertising, realistic textures, deadpan visual comedy, no readable text, no logos.</text:Value><text:Value id="broll-trophy-prompt">- A photograph with the texture of real iPhone footage. Generate a carefully framed composition, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. A surreal comedy photograph of a monumentally oversized championship trophy planted at the center spot of a real football stadium, towering over the goals and stands. Tiny groundskeepers and players stare up at it in disbelief. Gold metal, dramatic stadium floodlights, realistic grass, cinematic sports photography, ridiculous scale, no readable text, no logos.+ A surreal comedy photograph of a monumentally oversized championship trophy planted at the center spot of a real football stadium, towering over the goals and stands. Tiny groundskeepers and players stare up at it in disbelief. Gold metal, dramatic stadium floodlights, realistic grass, cinematic sports photography, ridiculous scale, no readable text, no logos.</text:Value><text:Value id="broll-greek-god-prompt">- A photograph with the texture of real iPhone footage. Generate a carefully framed composition, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. A photorealistic museum-style photograph of a muscular ancient Greek plaster statue posing like an overly confident fashion influencer: chest out, chin raised, one hand on his hip, the other adjusting imaginary sunglasses. Classical curls, white gypsum surface, heroic anatomy, dramatic gallery lighting, straight-faced visual comedy, no text.+ A photorealistic museum-style photograph of a muscular ancient Greek plaster statue posing like an overly confident fashion influencer: chest out, chin raised, one hand on his hip, the other adjusting imaginary sunglasses. Classical curls, white gypsum surface, heroic anatomy, dramatic gallery lighting, straight-faced visual comedy, no text.</text:Value><text:Value id="broll-greek-tragedy-prompt">- A photograph with the texture of real iPhone footage. Generate a carefully framed composition, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. A photorealistic production still from a classical Greek tragedy performed in an old opera house. A male actor in ancient robes kneels at center stage with his face buried in both hands, sobbing theatrically. A chorus of mourners surrounds him with raised hands, bowed heads, and black drapery. Painted temple scenery, smoky footlights, grand melodrama, emotionally excessive but historically theatrical, no modern objects, no text.+ A photorealistic production still from a classical Greek tragedy performed in an old opera house. A male actor in ancient robes kneels at center stage with his face buried in both hands, sobbing theatrically. A chorus of mourners surrounds him with raised hands, bowed heads, and black drapery. Painted temple scenery, smoky footlights, grand melodrama, emotionally excessive but historically theatrical, no modern objects, no text.</text:Value><text:Value id="broll-penaldo-social-prompt">A tightly cropped screenshot of a fictional social-media profile interface in the familiar clean Instagram visual language. The profile name must read exactly "Penaldo". The statistics must clearly show exactly "4,122 posts" and "679 million followers". The circular profile picture shows only the back silhouette of a footballer striking a penalty kick, one leg extended toward the ball. Crisp authentic mobile UI, white background, black interface text, blue verification badge, polished parody, no other names or numbers.</text:Value><text:Value id="broll-tired-dad-trophy-prompt">- A photograph with the texture of real iPhone footage. Generate a carefully framed composition, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. A photorealistic sports comedy photograph of a chubby, heavily bearded middle-aged uncle in a blue-and-white striped football jersey, panting hard and drenched in sweat while clutching the FIFA World Cup trophy against his chest. His cheeks are flushed, his eyes look exhausted yet reverent, and his belly presses against the jersey. Stadium grass and celebratory confetti, affectionate absurdity, realistic human anatomy, no text.+ A photorealistic sports comedy photograph of a chubby, heavily bearded middle-aged uncle in a blue-and-white striped football jersey, panting hard and drenched in sweat while clutching the FIFA World Cup trophy against his chest. His cheeks are flushed, his eyes look exhausted yet reverent, and his belly presses against the jersey. Stadium grass and celebratory confetti, affectionate absurdity, realistic human anatomy, no text.</text:Value><text:Value id="broll-walks-around-prompt">- A photograph with the texture of real iPhone footage. Generate a carefully framed composition, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. A photorealistic first-person POV from someone lounging on the grass of a football pitch with legs crossed. The lower edge of the viewer's own blue-and-white striped jersey is visible, along with shorts, knees, and cleats. Far away, two teams are playing an intense match: one team in blue-and-white shirts, the other in red. The relaxed spectator posture contrasts absurdly with the frantic game, natural daylight, realistic depth, no text.+ A photorealistic first-person POV from someone lounging on the grass of a football pitch with legs crossed. The lower edge of the viewer's own blue-and-white striped jersey is visible, along with shorts, knees, and cleats. Far away, two teams are playing an intense match: one team in blue-and-white shirts, the other in red. The relaxed spectator posture contrasts absurdly with the frantic game, natural daylight, realistic depth, no text.</text:Value><text:Value id="broll-hat-trick-prompt">- A photograph with the texture of real iPhone footage. Generate a carefully framed composition, as one frame cut out of video actually shot on an iPhone: genuinely real rather than glossy, carrying the texture of video and not of a posed photograph. The background stays clearly visible, with no depth-of-field blur. Skin texture is fine and real, the light is natural, and no part of the picture is broken. A surreal visual pun rendered as a polished studio photograph: one classic black-and-white football on a small pedestal, wearing three distinct elegant black top hats perched on it at playful angles. Dramatic spotlight, rich shadows, immaculate textures, whimsical deadpan comedy, an unmistakable hat-trick joke, no text, no logos.+ A surreal visual pun rendered as a polished studio photograph: one classic black-and-white football on a small pedestal, wearing three distinct elegant black top hats perched on it at playful angles. Dramatic spotlight, rich shadows, immaculate textures, whimsical deadpan comedy, an unmistakable hat-trick joke, no text, no logos.</text:Value><gpt:Image id="broll-hair-gel" prompt={broll-hair-gel-prompt}@@ -200,24 +221,6 @@<whisperx:SemanticTake id="messi-semantic" narrative={story}segment={story.segment.messi} media={messi-media.media} language="en"/>- <!-- Preview candidate path: the authored presenter image is formally encoded- as a silent MP4, normalized through the same waist, then given an honest- syllable-weighted estimated alignment over its complete frame count. -->- <pipeline:StillVideo id="ronaldo-still" source={presenter.image}- duration="10" clock={clock}/>- <pipeline:StillVideo id="messi-still" source={presenter.image}- duration="9" clock={clock}/>- <pipeline:Normalize id="ronaldo-preview-media" source={ronaldo-still.video}- video="primary-moving" audio="none" span-authority="video" clock={clock}/>- <pipeline:Normalize id="messi-preview-media" source={messi-still.video}- video="primary-moving" audio="none" span-authority="video" clock={clock}/>- <estimated:SemanticTake id="ronaldo-preview" narrative={story}- segment={story.segment.ronaldo} media={ronaldo-preview-media.media}- language="en" pace="normal" min="4" max="15" rounding="round"/>- <estimated:SemanticTake id="messi-preview" narrative={story}- segment={story.segment.messi} media={messi-preview-media.media}- language="en" pace="normal" min="4" max="15" rounding="round"/>-<speech:Track id="speech" visual-frame={speech-frame}visual-appearance={recipes.speech.visual} visual-z="0"><speech:Take source={ronaldo-semantic.take}/>@@ -237,19 +240,19 @@during="program" style={tier-style}appear-sound={ranking-appear-media.media}move-sound={ranking-move-media.media}>- <ranking:TierItem id="yamal" tier="s" preset="true" icon={yamal-icon.image}/>- <ranking:TierItem id="haaland" tier="a" preset="true" icon={haaland-icon.image}/>- <ranking:TierItem id="de-bruyne" tier="b" preset="true" icon={de-bruyne-icon.image}/>- <ranking:TierItem id="bellingham" tier="b" preset="true" icon={bellingham-icon.image}/>- <ranking:TierItem id="kane" tier="c" preset="true" icon={kane-icon.image}/>- <ranking:TierItem id="modric" tier="c" preset="true" icon={modric-icon.image}/>- <ranking:TierItem id="mbappe" tier="d" preset="true" icon={mbappe-icon.image}/>- <ranking:TierItem id="neymar" tier="d" preset="true" icon={neymar-icon.image}/>- <ranking:TierItem id="benzema" tier="d" preset="true" icon={benzema-icon.image}/>+ <ranking:TierItem id="yamal" tier="s" preset="true" icon={yamal-icon}/>+ <ranking:TierItem id="haaland" tier="a" preset="true" icon={haaland-icon}/>+ <ranking:TierItem id="de-bruyne" tier="b" preset="true" icon={de-bruyne-icon}/>+ <ranking:TierItem id="bellingham" tier="b" preset="true" icon={bellingham-icon}/>+ <ranking:TierItem id="kane" tier="c" preset="true" icon={kane-icon}/>+ <ranking:TierItem id="modric" tier="c" preset="true" icon={modric-icon}/>+ <ranking:TierItem id="mbappe" tier="d" preset="true" icon={mbappe-icon}/>+ <ranking:TierItem id="neymar" tier="d" preset="true" icon={neymar-icon}/>+ <ranking:TierItem id="benzema" tier="d" preset="true" icon={benzema-icon}/><ranking:TierItem id="ronaldo" tier="d" entry="drop"- icon={ronaldo-icon.image} during={story.segment.ronaldo}/>+ icon={ronaldo-icon} during={story.segment.ronaldo}/><ranking:TierItem id="messi" tier="s" entry="drop"- icon={messi-icon.image} during={story.segment.messi}/>+ icon={messi-icon} during={story.segment.messi}/></ranking:TierBoard><!-- Every Item consumes a projected Selection window. Later overlappingdiff --git a/packages/audio-track/README.md b/packages/audio-track/README.mdindex cd3bdc38..46484405 100644--- a/packages/audio-track/README.md+++ b/packages/audio-track/README.md@@ -9,3 +9,32 @@ pitch-preserving stretch occupancy, and lowers independent items to one ordinaryThe public audio waist uses one exact 48 kHz sample mapping. It contains no Provider, queue, path,Narrative identity, privileged speech lane, automatic extraction, loudness normalization or hiddenmixing behavior. Local and remote media Endpoints execute the same `AudioProgramPlan`.++## Authoring clips++The current Markup Track requires `semantic`; its Fragment derives ProgramSpace from that input.+Each Clip accepts one shared temporal Window form: `during="program"`, a Selection or Segment,+a Moment paired with `for`, or explicit `start`/`end` expressions. The source must already contain+normalized audio. This excerpt assumes imports, Script, SemanticTrack, Clock and source media exist:++```svml+<pipeline:Normalize id="hit-media" source={hit-source}+ video="none" audio="default" span-authority="audio" clock={clock}/>+<audio:Track id="effects" semantic={speech.semantic}>+ <audio:Clip source={hit-media.media} at={story.moment.reveal} for="600ms"+ playback="once" gain="0.45" fade-in="0f" fade-out="3f"/>+</audio:Track>+```++The Track publishes `.program` and `.track`; include the latter as a peer Film Track. `gain` is a+linear multiplier, defaulting to `1`; fade durations default to `0f`. Trim boundaries are source+durations (`f`, integer `ms`, or fractional `s`), distinct from the Clip's destination Window.+`trim-end` is the exclusive source endpoint, not an amount to subtract from the tail. For example,+`trim-start="1s" trim-end="3s"` selects the source interval from one to three seconds.++`once` / `once-start` plays once from the Window start; `once-end` aligns to its end. `loop` /+`loop-start` and `loop-end` repeat with the corresponding alignment. `stretch` preserves pitch and+requires authored `min-rate` and `max-rate` bounds. Other playback modes reject those rate bounds.+Choose occupancy from the intended sound; the Track does not silently loop music or stretch a hit+to fill a longer Window. Its source level is retained unless gain or another explicit audio operation+changes it.diff --git a/packages/audio-track/src/manifest.ts b/packages/audio-track/src/manifest.tsindex 7edbd604..75db20f0 100644--- a/packages/audio-track/src/manifest.ts+++ b/packages/audio-track/src/manifest.ts@@ -91,9 +91,9 @@ export const audioTrackMarkupSurfaces = [{summary: "Selects the explicitly prepared audio this Clip plays." },...temporalWindowAttributeVocabulary,{ name: "trim-start", kind: "literal", required: false,- summary: "Removes an exact duration from the head of the source." },+ summary: "Sets the source start position; defaults to the source beginning." },{ name: "trim-end", kind: "literal", required: false,- summary: "Removes an exact duration from the tail of the source." },+ summary: "Sets the exclusive source end position; defaults to the source end, not a duration subtracted from its tail." },{ name: "playback", kind: "literal", required: false,values: ["once", "once-start", "once-end", "loop", "loop-start", "loop-end", "stretch"],summary: "Decides how the source occupies a window longer or shorter than itself; defaults to `once`." },@@ -120,8 +120,8 @@ export const audioTrackMarkupSurfaces = [{notes: ["A Track requires at least one Clip, and neither a Track nor a Clip accepts text content.","A Clip states exactly one window form: `during`, `at` with `for`, `until` with `for`, or `start` with `end`.",- "A point expression is `program.start`, `program.end`, `selection.start`, `selection.end` or `moment.cue`, each optionally offset by `+` or `-` and a duration, or a bare duration read as an absolute position.",- "`selection` and `moment` cannot be written together.",+ "A point expression is `program.start`, `program.end`, `selection.start`, `selection.end`, `segment.start`, `segment.end` or `moment.cue`, each optionally offset by `+` or `-` and a duration, or a bare duration read as an absolute position.",+ "Bind selection, segment and/or moment only when the start/end expressions use them; different endpoints can use different bindings.","`min-rate` and `max-rate` are rejected unless `playback` is `stretch`.",],},diff --git a/packages/caption-fine/README.md b/packages/caption-fine/README.mdindex 16701aa9..11ed8bb6 100644--- a/packages/caption-fine/README.md+++ b/packages/caption-fine/README.md@@ -12,7 +12,9 @@ does not carry a private visual role.```An optional Spatial Region Timeline replaces only the moving placement point. The timeline is authored-numeric input, normally measured from the face before the Build; it is not a face-tracking request:+numeric input, measured from the actual footage before the composition pass that consumes it;+it is not a face-tracking request. A prior Build may produce that footage with ordinary fixed Caption,+and a later Run can reuse the same media and alignment while changing placement:```xml<space:RegionTimeline id="heads" within={vertical} recipe={tracking.heads.default}/>@@ -20,7 +22,7 @@ numeric input, normally measured from the face before the Build; it is not a facsemantic={speech.semantic} program={captions} regions={heads}/>```-Every Cue must carry one Script Role. When the Region Timeline contains a measured region for that+With `regions`, every Cue must carry one Script Role. When the Region Timeline contains a measured region for thatRole and Frame, the Track places the Cue at the region's top center. When that Role has a Track butthe current Frame is `null`, the Cue is not rendered: absence of evidence never becomes a guessedposition. A Role with no Track uses the Style's authored `x` and `y`, so unrelated speakers remaindiff --git a/packages/caption/README.md b/packages/caption/README.mdindex f8859e1a..ceb1f9f3 100644--- a/packages/caption/README.md+++ b/packages/caption/README.md@@ -22,3 +22,37 @@ another timing source. They are part of the common Caption contract for structurread-only projection of token evidence, not an author-controlled placement window. It neverthelesscarries the exact `spaceId`, `narrativeId` and `documentId`, so another SemanticTrack or Script cannotbe substituted silently.++## Extending with another rendering family++The common public values are defined in [types.ts](src/types.ts), with exports in+[index.ts](src/index.ts). A family supplies:++- a Style decoder producing `CaptionStyleIntent` with its own `rendering.family` and validated+ `rendering.parameters`;+- an optional family schedule Type and deterministic scheduling Producer;+- a rendering Producer returning the ordinary `VisualTrack` Type;+- a Surface/Fragment connecting the common document, program and timing to those Producers.++`caption:Program` remains the common author Surface. `Use` selects exactly one of `role`, `selection`+or `attribute`; `Mute` selects a Role or Selection. Overlapping unit Style applications are resolved+in authored order, while word attributes produce separate `wordRuns`. This is explicit application,+not CSS inheritance. A family chooses whether and how it supports word-specific runs.++Reuse `captionProducers.temporalizeDocument` with inputs `document`, `semantic` and `program`, or the+corresponding exported timing Fragment. It emits a `TimedCaptionProjection` whose Cues contain+`styleId`, `startFrame`, `endFrameExclusive` and measured `units`. Each unit carries `unitId` and its+own boundaries. Recover display Words and their associations from the same `CaptionDocument`; do+not infer N:M timing by evenly dividing a Cue or splitting its rendered string.++Scheduling may derive visible intervals while preserving those measured boundaries. Use+`assertCaptionProgramForDocument`, `assertTimedCaptionProjection` and, for visible scheduling,+`applyCaptionMute` as appropriate. Retain `spaceId`, `narrativeId` and `documentId` in a schedule so+the renderer can check its inputs. The renderer should consume that schedule without silently+changing Cue grouping, lead or tail again.++`@hypit/caption-fine` provides a working example in its `src/fragment.ts`, `src/schedule.ts` and+`src/render.ts`: one shared timing join, then family schedule and render operations. A new structural+family replaces Fine's uniform-word layout and parameters, not the common speech truth. Its own+Manifest, activation and vocabulary belong to the project package; no change to this common package+is required merely to add a new visual family.diff --git a/packages/film/README.md b/packages/film/README.mdindex 429bf366..59b30924 100644--- a/packages/film/README.md+++ b/packages/film/README.md@@ -16,9 +16,30 @@ remains order-independent because the public contract canonicalizes Tracks and vbelongs to each Present's absolute stacking key.The official Structured Surface validates an imported generic SVS Recipe into a nominal-`FilmProgram`, receives CanvasSpace and ProgramSpace through separate explicit edges, type-checks+`FilmProgram`, receives CanvasSpace and SemanticTrack through separate explicit edges, projects+ProgramSpace from that SemanticTrack, type-checkseach `<film:Track source={...}/>` reference and generates the finite fold. CanvasSpace is the onlydimension truth, ProgramSpace is the only frame-rate truth, and FilmProgram owns only assemblyidentity and clear color.Child order is organizational: Track and Present identity, timing and absolute stacking remain intheir own typed values. Final rendering is a separate author package and capability.++For example, after the named inputs are declared:++```svml+<film:Film id="main" canvas={vertical} semantic={speech.semantic}+ appearance={recipes.film.vertical}>+ <film:Track source={speech.visual}/>+ <film:Track source={speech.audio}/>+ <film:Track source={coverage.visual}/>+ <film:Track source={captions.track}/>+ <film:Track source={music.track}/>+</film:Film>+```++The public output is `main.composition`. Including a visual output does not automatically include+its sibling audio output. An opaque upper layer can cover a performance while its audio continues;+reordering these children is not how an author changes that visual stacking.++The `appearance` Recipe uses `background`, for example `film.vertical { background: #18212A; }`.+The compiled FilmProgram's field is `clearColor`; it is not a Recipe key named `clear-color`.diff --git a/packages/markup/README.md b/packages/markup/README.mdindex a366db7c..e138f113 100644--- a/packages/markup/README.md+++ b/packages/markup/README.md@@ -6,7 +6,7 @@ the leading Import Prologue, namespace binding, generic structured elements andpackage-owned Markup Surface Host facets.The package has no built-in Script, media or video vocabulary. `<script>` is accepted only when an-the imported package contributes that tag through a locked Markup Host facet. The semantic Module+imported package contributes that tag through a locked Markup Host facet. The semantic ModuleManifest contains no XML or parser fact. The current registry is for trusted official/in-process use; it is not a securitysandbox for third-party parser code.@@ -22,6 +22,33 @@ sandbox for third-party parser code.</svml>```+## Author syntax++Imports must precede body declarations. `from="@scope/package@1"` selects a logical Module ABI from+an installed package; `source="./other.svs"` selects another SourceUnit. A Host may also resolve a+non-relative Source locator such as an explicitly exported package Source; this is the same Markup+construct and does not give Markup package-manager knowledge. An `as` alias prefixes that module's+tags or that Source's public bindings. A Source import is not textual inclusion and does not make+its private graph identifiers public. Aliasing does not rewrite authored Narrative+identities: separately declaring two Scripts named `story` in one closure is a conflict, even when+their binding paths use different import aliases.++Attributes are quoted strings or whole-value references such as `source={portrait.image}`. Braces+contain a binding path, not JavaScript. The receiving Surface owns literal number/boolean parsing,+expression units, allowed children and output names. There is no universal `.image` or `.track`+suffix. Generic structured text and quoted attributes decode `<`, `>`, `&`, `"` and+`'`; a raw Surface such as Script owns its own body grammar and escapes. Comments use+`<!-- ... -->` outside raw bodies.++Put values before structured Surfaces that resolve them. During decoding, `resolveReference` sees+imported public values, earlier authored Records and earlier Component output declarations. It can+inspect an inline Recipe immediately but cannot inspect a generated value that has not run.+Arbitrary forward references are not a general author-language guarantee. These declaration-order+requirements are distinct from execution order, which follows graph dependencies and demanded+Outputs.++## Frontend implementation+Decoding has two passes. `discoverMarkup` reads only the root and complete leading Import Prologue.After Driver supplies one immutable resolved closure, `decodeMarkup` freezes the visible Surfacescope and parses the body. A raw Surface receives the source cursor immediately after its opening@@ -42,9 +69,10 @@ that does not request assets remains unchanged apart from being awaitable.Markup validates source ranges, duplicate identities, Host-facet-declared Record types and completeFragment references. It strips diagnostic ranges before sealing one `hypit.author-module@1`, so-source reflow does not change author semantics. It never expands a Fragment or resolves a component-input while reading the body. After every declaration has been collected, `@hypit/elaborator`-resolves forward references and emits the Core Graph.+source reflow does not change author semantics. It does not expand Fragments while reading the+body. Surface handlers can resolve references through the scoped decoder context described above;+after all drafts are collected, `@hypit/elaborator` resolves their graph wiring and emits the Core+Graph. That later graph pass does not make future authored Records available to an earlier decoder.Direct `decodeMarkup()` calls require every source import to be supplied as an already resolvednamespace. The reference `compileSourceClosure()` orchestration in `@hypit/elaborator` recursivelydiff --git a/packages/media-track/README.md b/packages/media-track/README.mdindex 13de6ed9..d410b471 100644--- a/packages/media-track/README.md+++ b/packages/media-track/README.md@@ -2,8 +2,9 @@Official provider-free Media Item and Sequence authoring package.-`Track` accepts explicit ProgramSpace, CanvasSpace and, when semantic points are used, a-SemanticTrack. Each independently timed `Item` receives an explicit SpatialFrame and either+The domain computation consumes explicit ProgramSpace and CanvasSpace. The current Markup `Track`+requires `semantic` and `canvas`; its Fragment projects ProgramSpace from that SemanticTrack, even+when an Item uses clock expressions. Each independently timed `Item` receives an explicit SpatialFrame and eitherone direct source or ordered Paint/sample layers. A visual source is always named by its actualform: `image`, prepared `media`, or compositable `surface`. Moving video must be prepared by anexplicit Media Pipeline `Normalize` Operation before it enters the Track; the Track never guesses a@@ -23,3 +24,82 @@ restricted internal projection.The package lowers only track composition. It does not add Media, B-roll or Provider meaning to Core,Film, Composition or HyperFrames. “B-roll” is an editorial use of an ordinary Item or Sequence.++## Author Items and replacement Sequences++These excerpts assume the imports, named assets, normalized media, Frames, Extents, Script,+SemanticTrack and Recipes are already declared:++```svml+<media-track:Track id="coverage" semantic={speech.semantic} canvas={vertical}>+ <media-track:Item image={photo} extent={photo-extent} frame={full-frame}+ during={story.selection.example} appearance={recipes.media.still}/>+ <media-track:Item media={clip-media.media} frame={full-frame}+ during={story.selection.proof} appearance={recipes.media.clip}/>+</media-track:Track>+```++One Item selects exactly one direct source form (`image` + `extent`, prepared `media`, or `surface`)+or declares ordered `Paint`/`Layer` children. The current Surface has no raw `video` attribute.+Appearance owns stack order, fit, clipping and source sampling; a `motion` Recipe owns lifecycle+movement. `Sampling` children can move the sampled content inside the Frame independently.++```svml+<media-track:Track id="cards" semantic={speech.semantic} canvas={vertical}>+ <media-track:Sequence id="steps" frame={card-frame} appearance={recipes.media.card}+ until={story.selection.demo} until-boundary="end">+ <media-track:Member id="one" image={first} extent={card-extent}+ at={story.moment.first}/>+ <media-track:Member id="two" image={second} extent={card-extent}+ at={story.moment.second}/>+ <media-track:Handoff id="one-two" from="one" transition={recipes.transition.cards}/>+ </media-track:Sequence>+</media-track:Track>+```++Members have ordered activation Instants; the next activation ends the preceding logical phase,+and the Sequence has an explicit terminal Instant. Each adjacent pair needs a Handoff, including+an explicit `cut` when no blended transition is intended. `from` names the outgoing Member; its+successor is the incoming Member. A Handoff's duration and boundary ratio place the visual transition+around that logical boundary without moving the semantic event.++An Item or Member can select audio using `source-audio`: `content` for its direct source, or a child+Layer id for layered content. This is opt-in. `Sound` can instead attach normalized audio to Item+`at="enter"` / `at="exit"`, or to a Sequence `handoff="one-two"`. The Track publishes `.visual` and+`.program`, plus `.audio` when sound is present; Film must include the desired peer outputs.++## Item windows and source playback++The Item's projected Window determines when it is active in the program. A timed sample's Recipe+separately controls which source frames it occupies there. `playback` defaults to `once-start`:++| Value | Source occupancy inside the Window |+| --- | --- |+| `once-start` | Play at native rate from the source start. Truncate if the Window is shorter; draw no source after it ends if the Window is longer. |+| `once-end` | Align native playback to the Window end. Use the source tail if the Window is shorter; leave the initial excess unoccupied if longer. |+| `hold-start` | Play from the start, then hold the last source frame over any remaining Window. |+| `hold-end` | Align playback to the end, holding the first source frame over any initial excess. |+| `loop-start` / `loop-end` | Repeat at native rate, aligning the loop phase to the start or end. |+| `stretch` | Retime the selected source range across the full Window. |++`trim-start` and `trim-end` are a pair of integer source-frame boundaries, with an exclusive end.+Omit both to use the full source. The playback rules apply to the trimmed range when present.+For `stretch`, the source-frame advance is `sourceLength / targetLength`. For native playback, moving+media has already been normalized to ProgramSpace's frame rate, so one program frame advances one+source frame. These are source sampling choices; they do not move a Selection's semantic boundaries.++A two-second Item can use the beginning of a four-second generated clip without requesting a+model-invalid two-second generation. Conversely, an Item active longer than its source must choose+whether a return to the layer below, a hold, a loop or retiming is intended. No mode is universally+correct for B-roll. Extending a Window with `once-start` does not extend the underlying video.++Durationless still images refuse `playback`, `trim-start` and `trim-end`. Their spatial placement and+Item Window are sufficient. A static B-roll picture does not need conversion to a generated clip.++Visual and source-audio projections remain separate. Selecting source audio applies the corresponding+sampling policy to that audio; a held visual frame does not synthesize held sound. A silent B-roll+layer can cover a speaking Track while the latter's audio continues unchanged.++For adjacent semantic Items that should have no uncovered pause, author touching Selection boundaries+using Script affinity. Also inspect opacity transitions and source exhaustion: touching Windows alone+do not guarantee opaque visual coverage.diff --git a/packages/media-track/src/manifest.ts b/packages/media-track/src/manifest.tsindex 47aa46ec..76791c5e 100644--- a/packages/media-track/src/manifest.ts+++ b/packages/media-track/src/manifest.ts@@ -418,11 +418,11 @@ export const mediaTrackMarkupSurfaces = [{recipe: motionRecipeProperties },{ name: "clip", kind: "reference", required: false, accepts: [spatialTypes.path],summary: "Clips the Sequence to an authored Path." },- { name: "until", kind: "expression", required: true, values: ["program.end"],- accepts: [narrativeTypes.moment, narrativeTypes.selection],- summary: "Ends the Sequence at the literal `program.end`, at a Moment, or at a Selection." },+ { name: "until", kind: "reference", required: true,+ accepts: [narrativeTypes.moment, narrativeTypes.selection, narrativeTypes.excerpt],+ summary: "Ends the Sequence at a Moment or a chosen Selection/Segment boundary." },{ name: "until-boundary", kind: "literal", required: false, values: ["start", "end"],- summary: "Chooses which edge of the ending Selection ends the Sequence; defaults to `end`." },+ summary: "Chooses which edge of the ending Selection or Segment ends the Sequence; defaults to `end` and is refused for a Moment." },] },],ports: [@@ -442,7 +442,7 @@ export const mediaTrackMarkupSurfaces = [{</media-track:Track>`,notes: ["A Track requires at least one Item or Sequence and accepts no text content.",- "An Item states exactly one window form: `during`, `at` with `for`, `until` with `for`, or `start` with `end`; `selection`, `segment` and `moment` bind a start/end window and cannot be written together.",+ "An Item states exactly one window form: `during`, `at` with `for`, `until` with `for`, or `start` with `end`. Bind `selection`, `segment` and/or `moment` only when the start/end expressions use them; different endpoints can use different bindings.","A point expression is `program.start`, `program.end`, `selection.start`, `selection.end`, `segment.start`, `segment.end` or `moment.cue`, each optionally offset by `+` or `-` and a duration, or a bare duration read as an absolute position.","A unit that names a direct source names exactly one of `image`, `media` or `surface`; `extent` is required with `image` and refused otherwise.","`audio-gain` is refused without selected source audio.",@@ -452,16 +452,17 @@ export const mediaTrackMarkupSurfaces = [{"An `enter` or an `exit` operator requires its own `enter-frames` or `exit-frames`.","An Item written with a direct source accepts `<Sampling>` and `<Sound>` children; an Item written without one accepts `<Paint>`, `<Layer>` and `<Sound>` children, and requires at least one Paint or Layer.","A Sequence requires at least two `<Member>` children, exactly one `<Handoff>` for every adjacent Member pair, and accepts `<Sound>` children.",- "`until-boundary` is refused for `program.end` and for a Moment.",+ "Sequence `until` is a semantic reference, not a literal `program.end`. `until-boundary` defaults to end for a Selection or Segment and is refused for a Moment.","A Member is one picture in the replacement order:",["| Attribute | Kind | Required | Meaning |","|---|---|---|---|","| `id` | identifier | no | Names this Member; the Sequence derives `<sequence>.member.<index>` otherwise |",- "| `at` | reference (@hypit/narrative@1#NarrativeMoment, @hypit/narrative@1#NarrativeSelection) | yes | The Moment or Selection that activates this Member |",- "| `boundary` | literal (start, end) | with a Selection | Which edge of the activating Selection the Member starts on; refused for a Moment |",+ "| `at` | semantic reference | one timing form | A Moment, Selection or Segment that activates this Member |",+ "| `boundary` | literal (start, end) | with a Selection or Segment | Which edge activates this Member; refused for a Moment |",+ "| `instant` | expression | one timing form | Explicit point expression instead of `at`; bind selection, segment or moment if the expression uses it |","| `appearance` | reference (@hypit/svs@1#Recipe) | no | This Member's own Recipe in place of the Sequence's |",- "| `image` | reference (@hypit/artifact@1#Blob) | one source form | A durationless still image, which also requires `extent` |",+ "| `image` | reference (@hypit/artifact@1#BlobArtifact) | one source form | A durationless still image, which also requires `extent` |","| `media` | reference (@hypit/media@1#SynchronizedMedia) | one source form | An explicitly prepared timed source |","| `surface` | reference (@hypit/media@1#CompositableSurfaceRef) | one source form | An alpha-aware still or timed Surface |","| `extent` | reference (@hypit/spatial@1#IntrinsicExtent) | with `image` | The authored pixel extent of the still image |",@@ -475,7 +476,7 @@ export const mediaTrackMarkupSurfaces = [{"| Attribute | Kind | Required | Meaning |","|---|---|---|---|","| `id` | identifier | no | Names this Handoff; the Sequence derives `<sequence>.handoff.<index>` otherwise |",- "| `from` | literal | yes | The id of the outgoing Member, which must be the Member preceding this Handoff |",+ "| `from` | literal | yes | The outgoing Member id; Handoffs are ordered by adjacent Member pair |","| `transition` | reference (@hypit/svs@1#Recipe) | yes | The Recipe for the handoff operator, its length, its boundary and its audio treatment |",].join("\n"),"A `transition` Recipe carries exactly these properties:",@@ -507,7 +508,7 @@ export const mediaTrackMarkupSurfaces = [{"| Attribute | Kind | Required | Meaning |","|---|---|---|---|","| `id` | identifier | no | Names this layer, which `source-audio` selects by; the Track derives one from the unit and the layer position otherwise |",- "| `image` | reference (@hypit/artifact@1#Blob) | one source form | A durationless still image, which also requires `extent` |",+ "| `image` | reference (@hypit/artifact@1#BlobArtifact) | one source form | A durationless still image, which also requires `extent` |","| `media` | reference (@hypit/media@1#SynchronizedMedia) | one source form | An explicitly prepared timed source |","| `surface` | reference (@hypit/media@1#CompositableSurfaceRef) | one source form | An alpha-aware still or timed Surface |","| `extent` | reference (@hypit/spatial@1#IntrinsicExtent) | with `image` | The authored pixel extent of the still image |",@@ -536,7 +537,7 @@ export const mediaTrackMarkupSurfaces = [{"| `gain` | literal | no | Linear gain applied to this sound; defaults to `1` |",].join("\n"),"A Sound states exactly one trigger form, and `handoff` only names a Handoff of the Sequence it is written in.",- "A direct source is sampled into a layer named `content`, so `audio=\"include\"` on a direct video selects that audio itself and refuses `source-audio` beside it.",+ "A direct prepared source is sampled into a layer named `content`; use `source-audio=\"content\"` to include its audio. Layered units select a Layer id. Raw `video` and `audio=\"include\"` attributes are not accepted.",],},}] as const;diff --git a/packages/run-markup/README.md b/packages/run-markup/README.mdindex e86ba39d..cf3ba589 100644--- a/packages/run-markup/README.md+++ b/packages/run-markup/README.md@@ -18,7 +18,9 @@ One Run Source is self-described and names one Author Source explicitly:The Author Source's own Header—not this file and not its suffix—selects its Author Frontend.-`<value>` declares a typed zero-input `StoredValue`. `<file>` admits ordinary source+`<value>` declares a typed zero-input `StoredValue`; its JSON file includes the wrapper, such as+`{"kind":"inline","value":{"enabled":true}}`, with the inner value matching the declared Type.+`<file>` admits ordinary sourcebytes as the explicitly named blob Type; Run itself does not assume an Artifact module.`<build-record>` exposes one exact named Output from a prior Build Result as a zero-input Candidate. `<fragment>`instantiates a trusted package Fragment and may@@ -31,6 +33,28 @@ unreached local values, local files and historical Outputs are not opened. Coreand the historical value receives no automatic semantic-relationship claimagainst the output it is selected to satisfy.+Results are not selected automatically by repeating a Run, retaining an output name, or changing only+its Targets. Cross-Build reuse requires an explicit Candidate selection. Removing a `satisfy` returns+that Logical Output to its primary Candidate and can make the displaced generation route reachable+again. Preserve unrelated selections when revising a Run.+Imports form the opening prologue. They name Fragment libraries already admitted by the Host; theydo not install packages. Provider credentials, queues, Stores, Endpoint bindings and inlinecallbacks are not Run language elements.++## Fragment input references++`<input name="media" from="opening-media.media"/>` names a public value of the Run's Author+entry. A computed Author Output remains a graph dependency, with its selected Candidate applied.+The reference is not another Run Candidate name and does not expose a nested Source namespace that+the Author entry has not published.++`<input name="duration" value="5"/>` supplies a scalar. The parser recognizes numbers, booleans+and null; other text is a string. Typed object inputs are obtained through `from`, not by placing+JSON in a scalar attribute. `<value>` reads a serialized typed value as a Candidate for a Logical+Output; it does not introduce an Author binding for Fragment inputs.++Fragment exports are available as `<instance-id>.<export-name>`. Optional `<export name="video"/>`+children expose selected exports; omitting them exposes all Fragment exports. Their shared required+operations remain one instance. Imports precede Targets and Candidate declarations, following the+required `author` declaration.diff --git a/packages/script/README.md b/packages/script/README.mdindex d41f55d3..4d8254f9 100644--- a/packages/script/README.md+++ b/packages/script/README.md@@ -13,8 +13,8 @@ know that Script, Segment or Narrative exist.@answer<opening>- <ALICE> 我先说第一句话。- <BOB> 然后我来回答。+ <ALICE> I will make the first point.+ <BOB> Then I will answer.</opening><pause/>@@ -25,11 +25,16 @@ know that Script, Segment or Narrative exist.`<opening>` opens a Segment named `opening`; `</opening>` closes that exact Segment. While theparser is inside a Segment, a valid bare tag such as `<ALICE>` is a Role Cue. This is parser state,not indentation: the compact spelling-`<opening><ALICE>我先说。<BOB>我回答。</opening>` has the same semantic value.+`<opening><ALICE>I speak first.<BOB>I answer.</opening>` has the same semantic value.A Role Cue is optional. Text before the first Role Cue is a roleless Turn, and Role state is resetwhen every Segment closes; a Role can never leak into the following Segment.+An empty Segment such as `<empty></empty>` (or `<empty/>`) is valid. `empty` is an ordinary+author-chosen name, not a reserved keyword. It retains the Segment identity and both boundary anchors+while contributing no Tokens or spoken text. This supports wordless passages in the same semantic+model: the associated normalized media, not the empty tag, determines the SemanticTake's duration.+The package exports its Manifest, `parseScript`, semantic/source-map projection helpers, asemantic-preserving formatter and the raw `decodeScriptSurface` handler. Source ranges and parserstate remain private to Script; its authored Narrative Record uses the Frontend-neutral type from@@ -44,7 +49,8 @@ The Surface exports one full Narrative plus narrow, immutable views:- `script.segment.<id>.speech` is ordinary pronunciation-only `Text` for duration estimation or TTS;- `script.caption` is one complete `CaptionDocument`: ordered display Words, N:M Alignment Units,and authored Cue Breaks;-- `script.selection.<id>` is a reusable explicit Selection.+- `script.selection.<id>` is a reusable explicit Selection;+- `script.moment.<id>` is a reusable explicit Moment.`@hypit/caption` projects `script.selection.<id>` or a Role onto complete Caption Alignment Units;it then joins those units to a SemanticTrack for frame timing. Seedance consumes dialogue `Text`,@@ -66,6 +72,58 @@ CaptionDocument contracts.`really{emphasis,keyword}`. It becomes `CaptionDisplayWord.attributes`; it is not a Selection anddoes not carry timing.-An unescaped `@` is legal only on the spoken side of Dual Text. The display side is literal; write+Within Dual Text, an unescaped `@` is legal only on the spoken side. The display side is literal; write`\@` if an at-sign must be shown. `||` is an authored Caption Cue Break and must occur betweencomplete Alignment Units.++## Segments, turns and Cues are different boundaries++A Segment names a structural production passage. It can contain several Role turns and be performed+by one Take with several edited shots. A Role Cue changes who speaks; it neither creates a character+asset nor requires another generated Take. `||` changes Caption grouping between complete Alignment+Units; it does not split the Segment, cut the picture or end a Selection.++```svml+<script id="story">+ <exchange>+ <HOST> I use it || every day, || since <2012 | twenty twelve>.+ <GUEST> Even @proof on holiday @/proof?+ <HOST> @answer! Especially then.+ </exchange>+</script>+```++Dual Text preserves display spelling while supplying an explicit pronunciation. Its N:M Alignment+Unit is indivisible for Caption timing and authored Cue Breaks. Roles are lexical speaking cues;+the Source's model references and action direction bind them to the intended performers.++## Selection and Moment affinity++A marker selects the adjacent semantic anchor; it does not write a timecode. Inside a spoken+passage, its affinity normally chooses a neighboring word boundary:++| Marker | Boundary |+| --- | --- |+| `@name` | Selection opens at the next word's start: right affinity. |+| `~@name` | Selection opens at the previous word's end: left affinity. |+| `@/name` | Selection closes at the previous word's end: left affinity. |+| `@/name~` | Selection closes at the next word's start: right affinity. |+| `@name!` | Moment at the next word's start: right affinity. |+| `~@name!` | Moment at the previous word's end: left affinity. |++At structural edges, the parser retains the corresponding Segment or program boundary rather than+inventing a neighboring word. Selection and Moment ids share one namespace. Selections can overlap,+cross and span Segments; unlike tags, they do not need to nest.++To join two visual Selections without exposing their inter-word pause, match affinity on both sides:++```text+@coffee my coffee @/coffee ~@smoothie my smoothie @/smoothie+@coffee my coffee @/coffee~ @smoothie my smoothie @/smoothie+```++These are alternative spellings, not two occurrences to put in the same Script. In the first,+“coffee” ends both the first Window and the gap's left boundary, so the smoothie Selection owns the+pause. In the second, the next “my” starts both touching boundaries, so the coffee Selection owns it.+Plain `@/coffee @smoothie` leaves the gap between previous word end and next word start outside both.+The media consumer still decides playback and visual coverage inside those projected Windows.diff --git a/packages/seedance-kits/README.md b/packages/seedance-kits/README.mdindex d1900d2d..4240cb2d 100644--- a/packages/seedance-kits/README.md+++ b/packages/seedance-kits/README.md@@ -21,7 +21,7 @@ ordinary explicit graph edges:</text:Render><seedance:ReferenceVideo id="broll" model="mini" prompt={broll-prompt}- duration={broll-duration.duration} resolution="720p" aspect-ratio="9:16">+ duration="5" resolution="720p" aspect-ratio="9:16"><seedance:Reference image={scene}/><seedance:Reference image={product}/></seedance:ReferenceVideo>@@ -41,7 +41,7 @@ Speaker uses the same graph vocabulary. The Kit assumes `@image1` is the visible</text:Render><seedance:ReferenceVideo id="hook-take" model="mini"- prompt={hook-prompt} duration={hook-duration.duration}+ prompt={hook-prompt} duration="8"resolution="720p" aspect-ratio="9:16" generate-audio="true"><seedance:Reference image={presenter}/><seedance:Reference audio={voice}/>@@ -54,10 +54,82 @@ The seven templates are:- `broll-v1`: silent visual micro-story;- `podcast-v1`: two fixed podcast views with two voices;- `call-v1`: two live video-call reverse views;-- `street-interview-v1`: interviewer, guest and shared street views, two voices and microphone handoff;+- `street-interview-v1`: interviewer, guest and shared street views, two voices and microphone positioning;- `motion-reference-v1`: preserve the subject and transfer body motion only;- `camera-reference-v1`: preserve the subject and transfer camera language only.-Speaker, B-roll, Podcast, Call and Street Interview expose finite orthogonal prompt axes through-their Text Templates. Speed, trim, last-frame extraction and audio extraction are ordinary media-operations and intentionally do not live in these templates.+Speaker, B-roll, Podcast, Call and Street Interview expose prompt choices through their Text+Templates. The examples above use author-chosen literal durations. Measure adopted speech with+`hypit measure` first, then choose a duration within the selected model's declared range; there is+no duration-estimation node on the generation route.++## Recipe, dialogue and action++The Recipe carries recurring choices exposed by the selected template. `dialogue` receives the+Script Segment's `.dialogue` Text, retaining pronunciation and Role turns. Optional `action` supplies+this passage's attitude, attention, interaction and any motivated cuts admitted by the Kit.+Do not retype the complete spoken text into action or treat each Role turn as a required new Take.+For B-roll, `story` carries the silent visual events rather than spoken dialogue.++These files are project-vendored authoring material. Adapt a project copy when the production needs+a different prompt structure; a Kit is not a Core restriction or a hidden media-generation wrapper.+Reference order, duration, audio generation and output aspect ratio remain explicit on the model+Surface. Setting a prompt option cannot create a media reference or execute a postprocess.++## Speaker: edited UGC from one useful view++Reuse the same person-and-scene image and voice for ordinary Takes belonging to one UGC performance.+Each Take can begin from that shared reference and meet the others at a natural cut. There is no+required previous-tail-frame chain. Independent generation follows the intended edited form.++For example, a lively ranking host can use:++```svs+speaker.host {+ composition-stability: strict-locked;+ camera-motion: none;+ edit-rhythm: pause-trim-jump-cuts;+ performance: high-energy-ugc;+ gesture: expressive;+}+```++`edit-rhythm: pause-trim-jump-cuts` asks the model for an edited pause-trim rhythm at phrase boundaries.+It is compatible with a fixed camera and active performance. `continuous-take` is another authored+choice, not a universal requirement. Actual speed changes, trimming or audio extraction are explicit+media operations; selecting this Recipe does not perform them on the returned video.++## Podcast: two views and living listeners++`@image1` is Host A's view, `@image2` is Host B's view; `@audio1` and `@audio2` correspond to those hosts.+If Script uses names such as GIRL and GUY, state their correspondence to A/B in action. One Segment+can contain the entire short exchange. Choose speaker or reaction cuts for the conversation rather+than splitting it into a generation per line.++Derive a complementary host view from the first useful scene image. Product variations normally+derive from each host's own view plus the product reference. Action can state the meaningful handoff+and resulting possession. Silent listeners can adjust posture, glance down and look back up; they+must not speak the partner's lines. A split-screen opening has a different layout from this two-view+template and can use its own explicit prompt and combined reference image.++## Street interview: three supplied setups++The image order is interviewer A, guest B, shared two-person setup; the voice order is A, then B.+The interviewer always holds the microphone, moving it between speaking positions. The guest does+not take ownership merely because the microphone is extended toward them.++Action chooses cuts among the supplied setups. Guest answers can favor the guest close view, brief+neutral questions the shared view, and a meaningful interviewer reaction their close view. Small+opening and closing actions can establish an encounter: the guest looks up from another activity+as the interviewer approaches, and begins to leave after the final answer. These are examples of+motivated direction, not mandatory actions on every request.++## Keep gesture and scene direction readable++Use attitude and a few decisive actions. Prefer an explanatory hand movement or emphatic gesture+over exact numerical finger arrangements; speech or downstream graphics can state the number.+Avoid exaggerated reactions or constantly moving every person simply to prevent stillness.++For a multi-scene B-roll montage, `story` can name the references in scene order and give each a small+action. Choose an edit language compatible with cuts. “Continuous within each scene” and “one+continuous shot for the whole montage” ask for different results.diff --git a/packages/semantic-take-estimate/README.md b/packages/semantic-take-estimate/README.mdindex af8f7a8c..0829802b 100644--- a/packages/semantic-take-estimate/README.md+++ b/packages/semantic-take-estimate/README.md@@ -38,5 +38,38 @@ estimated Build product:<satisfy output="opening-semantic.take" candidate="opening-preview"/>```-This is ordinary Run candidate selection. There is no preview-specific `SemanticTake` subtype and-Studio still opens only after every required output has an actual candidate with no unresolved Need.+This is ordinary Run candidate selection. Studio can evaluate deterministic work and the immediate+capabilities that the selected Provider permits for transient execution. Other required media work+can be completed through a Build and selected from its Result.++## Run Fragment++The package also registers `@hypit/semantic-take-estimate@1#semantic-take` as a Run Fragment:++| Input | Type |+| --- | --- |+| `narrative` | `@hypit/narrative@1#Narrative` |+| `segment` | `@hypit/narrative@1#NarrativeExcerpt` |+| `media` | `@hypit/media@1#SynchronizedMedia` |+| `policy` | `@hypit/estimate@1#SpeechEstimatePolicy` |++Its export is `take`, an ordinary SemanticTake. The policy is a typed value with `language`,+`pace` or `rate`, `minimumSec`, `maximumSec`, `rounding` and optional `paddingSec`.+The Author Surface publishes its authored policy as `<id>.policy`. For an Author entry containing+the `opening-estimated` declaration above, a Run can use:++```svrun+<import as="estimate" from="@hypit/semantic-take-estimate@1"/>+<fragment id="timing" using="estimate:semantic-take">+ <input name="narrative" from="story"/>+ <input name="segment" from="story.segment.opening"/>+ <input name="media" from="opening-media.media"/>+ <input name="policy" from="opening-estimated.policy"/>+</fragment>+<satisfy output="opening-semantic.take" candidate="timing.take"/>+```++This selects estimated timing for the measured Take while retaining the required prepared-media+dependency and any Run choice for its source bytes. The estimate allocates the media's complete+frame count; it neither generates nor normalizes that media. It requires spoken Tokens. A wordless+Segment uses actual media boundaries directly when materializing its SemanticTake.diff --git a/packages/semantic-track/README.md b/packages/semantic-track/README.mdindex 9438af0d..0845fce3 100644--- a/packages/semantic-track/README.md+++ b/packages/semantic-track/README.md@@ -3,7 +3,9 @@`SemanticTrack` is the single global semantic-timeline fact. It is the continuous program skeleton:every item is a real segment-local `SemanticTake`, and global frame positions are the prefix sum oftheir normalized local frame counts. There are no gap items or empty items. A Take may carry only-audio or only visual media; only its word array may be empty. Visual placement is deliberately not+audio or only visual media; its word array may be empty. An empty Script Segment such as+`<empty></empty>` can therefore retain a real media span and both Segment anchors without speech.+An empty word array is not an empty Track item. Visual placement is deliberately notpart of this semantic truth.The package exposes pure projections to `ProgramSpace` and `AudioTrack`, plus thediff --git a/packages/spatial/README.md b/packages/spatial/README.mdindex 1c93760a..bb6fefb0 100644--- a/packages/spatial/README.md+++ b/packages/spatial/README.md@@ -9,10 +9,11 @@ The package exposes self-described `Canvas`, `Point`, `Path`, `Extent`, `RegionT`AnchoredFrame` and `AspectFrame` author Surfaces plus pure geometry functions and fixed-port Producers.`RegionTimeline` accepts already measured data rather than running a detector. Face detection and-tracking belong to reference observation before authoring: inspect the footage, keep the useful boxes-as ordinary numbers, and pass that finished timeline to Caption. A Build never starts a detector to-discover its own layout. One SVS Recipe holds the exact ProgramSpace frame count and named tracks whose-array positions are Frames:+tracking can observe the footage produced by an earlier Build: keep useful boxes as ordinary numbers,+then reuse that media while rendering Caption with the finished timeline. Measurement precedes the+composition that consumes it; it need not precede the first media-producing Build. RegionTimeline+does not start a detector to discover its own layout. One SVS Recipe holds the exact ProgramSpace+frame count and named tracks whose array positions are Frames:```svsheads.default {@@ -30,3 +31,10 @@ heads.default {Every measured region is normalized `[x, y, width, height]`; `null` says that this track has nomeasured region on that Frame. The Surface converts measured regions to the selected Canvas's pixelsat author time; it never fills missing Frames or interprets a track id.++Prepare measurements against the exact edited media, clock and Canvas used by the composition.+Individual Take measurements need their actual frame offsets in the program; boxes measured before+a crop, resize or inset need that placement transform. Face-to-head expansion, identity association,+cut handling and any interpolation are explicit external preparation decisions. The resulting+Recipe records their output. Caption consumers can assign Role meaning to track ids without making+Spatial aware of speakers, detector APIs or generated-media Providers.diff --git a/packages/stand-in/README.md b/packages/stand-in/README.mdindex a96b47ec..d3ebfd84 100644--- a/packages/stand-in/README.md+++ b/packages/stand-in/README.md@@ -21,3 +21,18 @@ materializes their deterministic files. Model packages do not know these FragmenRemove the `satisfy` line to select the Source's primary Candidate again. The card carries adiagonal STAND-IN watermark, its kind, frame and duration and — for video — a running timecode andprogress bar, so a cut that lands on the wrong frame remains visible.++## Fragment inputs and exports++| Fragment | Inputs | Export |+| --- | --- | --- |+| `image` | `canvas`: CanvasSpace | `image`: BlobArtifact |+| `video` | `canvas`: CanvasSpace, `duration`: SpeechDuration in seconds, `clock`: ProgramClock | `video`: BlobArtifact |+| `silence` | `duration`: SpeechDuration in seconds | `audio`: BlobArtifact |++`from` on a Run input refers to a public value of the Author entry; literal duration can use+`value="5"`. The card/silence capability still needs a selected media Provider. Studio can use it+when that Provider admits it for transient execution. An estimated SemanticTake is a separate+choice supplied by `@hypit/semantic-take-estimate` after normalization.++These are Run Fragments, so this package has no Markup Surface entries in `hypit vocabulary`.diff --git a/packages/svs/README.md b/packages/svs/README.mdindex 64442fe2..d0d32708 100644--- a/packages/svs/README.md+++ b/packages/svs/README.md@@ -14,11 +14,24 @@ JSON array/object syntax. SVS only parses and serializes those shapes. Theconsuming domain package owns their schema and meaning.```svs+<?svml using="@hypit/svs@1"?>+<sheet version="1">ranking.column {rank-colors: ["#ff3f56", "#ffa72d", "#eadc2a"];}+</sheet>```Source import aliases are owned by Source Closure compilation. A rule `caption.alice` remains thatrelative public export; importing it as `studio` exposes `studio.caption.alice` without changing theRecipe Record identity.++Each property ends with `;`. Duplicate properties in one rule and duplicate rule paths are errors.+Bare words such as `cover` remain strings; use quoted strings when punctuation could conflict with+the rule syntax. JSON arrays and objects require JSON quoting; they do not evaluate expressions or+graph references. Units such as `8f` or `50%` remain string values until a consuming package interprets+them. A consumer that needs graph media or font inputs takes explicit Source references separately.++The Recipe Source's rule paths do not imply parent/child inheritance: `caption.base` and+`caption.guest` are independent values. Any parameter defaults or overrides come from the declared+consumer, such as TextTemplate's defaults/Recipe/Param precedence, not from SVS itself.diff --git a/packages/typography-track/README.md b/packages/typography-track/README.mdindex a778b91a..22462c1b 100644--- a/packages/typography-track/README.md+++ b/packages/typography-track/README.md@@ -13,9 +13,10 @@ an explicit `SpatialFrame` input from `@hypit/spatial`; neither is hidden in theRecipe. The package exposes semantic typography rather than a rendering callback, arbitrary CSS orcross-Track access.-The provider-free `<text:Track>` Surface validates `.svs` Recipes, accepts Program, Selection,+The provider-free `<typo:Track>` Surface validates `.svs` Recipes, accepts Program, Selection,Moment or explicit point-expression timing plus explicit Point/Frame/Path edges, and produces the-same TypographyTrackProgram without changing Core or Film. `<text:Mask>` is a separate component that+same TypographyTrackProgram without changing Core or Film. The current Track Surface requires+`semantic` and derives ProgramSpace from it, including for explicit clock expressions. `<typo:Mask>` is a separate component thatconsumes one authored Text Program and one owned still Surface; advanced/multiline/Path masks failclosed and materialize through an independent package.@@ -25,10 +26,10 @@ Content has two explicit author forms:<import as="copy" from="@hypit/text@1"/><import as="typo" from="@hypit/typography-track@1"/>-<copy:Value id="headline">Intent, not timelines.</copy:Value>+<copy:Value id="headline-copy">A useful idea, clearly shown.</copy:Value><typo:Track id="titles" semantic={speech.semantic}>- <typo:Area id="headline" content={headline}+ <typo:Area id="headline" content={headline-copy}placement={layout.headline} style={title-style} during="program"/><typo:Area id="editorial" placement={layout.editorial}diff --git a/packages/typography-track/src/manifest.ts b/packages/typography-track/src/manifest.tsindex 6303c9bf..9cdfc332 100644--- a/packages/typography-track/src/manifest.ts+++ b/packages/typography-track/src/manifest.ts@@ -485,8 +485,8 @@ export const typographyTrackMarkupSurfaces = [</text:Track>`,notes: ["A Track requires at least one `<Point>`, `<Area>` or `<Path>` and accepts no text content of its own.",- "An item states exactly one window form: `during`, `at` with `for`, `until` with `for`, or `start` with `end`; `selection`, `segment` and `moment` bind a start/end window and cannot be written together.",- "A point expression is `program.start`, `program.end`, `selection.start`, `selection.end` or `moment.cue`, each optionally offset by `+` or `-` and a duration, or a bare duration read as an absolute position.",+ "An item states exactly one window form: `during`, `at` with `for`, `until` with `for`, or `start` with `end`. Bind selection, segment and/or moment only when the start/end expressions use them; different endpoints can use different bindings.",+ "A point expression is `program.start`, `program.end`, `selection.start`, `selection.end`, `segment.start`, `segment.end` or `moment.cue`, each optionally offset by `+` or `-` and a duration, or a bare duration read as an absolute position.","An item written without `content` owns its own document: direct text becomes one paragraph, and `<P>` children carry rich runs instead; a document mixes neither `<P>` children with direct text nor direct text with nested elements.","Neither a `<P>` nor a `<Span>` may be empty, and a `style` on either must name a Style Record authored in this Source or imported from another, because its typography and Paint are copied into the document as the item is decoded.",],diff --git a/packages/video-cli/README.md b/packages/video-cli/README.mdindex cfaf4d63..2f1be95e 100644--- a/packages/video-cli/README.md+++ b/packages/video-cli/README.md@@ -6,7 +6,10 @@ are activated only by explicit Source imports or Profile `use` entries.Every Frontend, Surface, deterministic Producer and Validator is activated from Source imports.Installing a new author package therefore does not require a video CLI or Core release. Source-imports never grant network, credential or process authority.+imports never grant network, credential or process authority. A data-only package can also export a+reusable Source directly, for example+`<import as="ugc" source="@hypit/gpt-image-kits/phone-ugc-v1"/>`; resolving that Source does not+activate package code.From any project: