Variables And Media

Prefer declarative bindings — no script needed for direct substitution:

• data-var-src="id" substitutes the element’s src (URL string or image {url}); the authored src is the fallback.
• data-var-text="id" substitutes the element’s own text; element children (nested clips, animated spans) are preserved.
• Every scalar variable is applied automatically as a --{id} CSS custom property on the composition root, so var(--id) CSS responds to overrides — no setProperty boilerplate.
• Bindings resolve identically in preview and render, and per-instance for sub-compositions.
• Caveat: media with audio should keep a real fallback src — render audio extraction reads the authored attribute (lint: media_variable_src_no_fallback).

For logic beyond direct substitution (loops, conditionals, derived values), read values once during initialization:

Variable Rules

• Supported types and their extra options (consumed by Studio’s editing UI):
  • string — optional placeholder, maxLength
  • number — optional min, max, step, unit
  • color — none
  • boolean — none
  • enum — required options: [{ "value": "...", "label": "..." }, ...]
• Always provide useful default values so preview works without CLI overrides.
• Use data-variable-values='{"title":"Pro"}' on sub-composition hosts for per-instance overrides.
• Use npx hyperframes render --variables '{"title":"Q4 Report"}' or --variables-file for render-time overrides.
• Add --strict-variables in CI: turns undeclared keys, type mismatches, and enum values not in options into errors instead of warnings.
• Read values once during init, not on every animation tick — variables don’t change mid-render.
• Media color grading can use exact variable references inside data-color-grading JSON. Use $gradingPreset or ${gradingIntensity} as the whole field value; the runtime resolves it from the current composition’s variables before applying shader adjustments, finishing details, blur/pixelate effects, and custom LUTs.

Two JSON Shapes (Easy to Confuse)

• data-composition-variables is an array of declarations (the schema): [{id, type, label, default}, ...]
• --variables and data-variable-values are objects keyed by id (the values): {"title":"Q4","accent":"#fff"}

Media

<video>/<audio> work at any nesting depth, including inside a sub-composition <template> or a wrapper <div>. The runtime discovers media with a flat document.querySelectorAll("video, audio"), resolves each element’s host composition via element.closest("[data-composition-id]"), and rebases its local data-start by the accumulated absolute start of every ancestor composition. A host at 2 with child media at 2 therefore starts that media at root time 4, consistently in preview, snapshot, extraction, and render. Legacy projects that deliberately authored a media start in root time must mark that element with data-hf-media-start-basis="global"; never infer the basis from overlapping numbers. New compositions should always use scene-local data-start. If a panel renders blank after a render, capture a per-frame snapshot and treat it as render-blocking.

The one real constraint is about timelines, not media placement: a sub-composition timeline cannot reach or animate host elements — neither document.querySelector("#host-id") nor a gsap selector string (tl.to("#host-id", …)) resolves across the boundary; a sub-comp timeline only drives its own subtree. So if a media element lives at the host root, its per-scene motion (scale/opacity/morph/tilt/breathing) must be authored on the MAIN timeline in index.html, at GLOBAL time (scene-local time + the scene slot’s data-start). Keeping the media inside the scene sub-comp instead lets that sub-comp’s own timeline animate it with scene-local time. For 3D tilt without a perspective parent, use gsap transformPerspective on the element. See composition-patterns.md archetype B.

Video elements must be muted and inline. Audio must be a separate <audio> element, even when it uses the same source file.

Media Rules

• Do not call video.play(), audio.play(), pause, or seek in composition code. HyperFrames owns playback.
• Do not drive host-root media from a sub-comp timeline: a sub-comp timeline cannot reach elements outside its subtree, so it has no effect. Drive host-root media from the main timeline at global time (or keep the media inside the sub-comp whose timeline animates it).
• Do not animate timed media element dimensions; animate a non-timed wrapper instead.
• Do not nest video inside a timed wrapper. lint rejects a <video data-start> whose ancestor also carries data-start (video_nested_in_timed_element, error), and the failure is real: the frame extractor resolves the video’s start from its own data-start without the wrapper’s offset, while visibility uses the wrapper’s window. The clip then shows the wrong source frames and disappears partway through its slot. Put the timing on the wrapper or on the media element, never both.
• Sub-compositions are exempt and work. A <video>/<audio> inside a sub-composition renders identically to one at the host root, because a composition host propagates its offset. Only plain timed wrappers (a <section data-start> around a <video data-start>) break.
• Never add crossorigin to <video>/<audio>. lint rejects it unconditionally (media_crossorigin_breaks_preview, error) because a media host without Access-Control-Allow-Origin then fails silently in preview while renders still work, hiding the bug. There is no suppression, so this holds even for the canvas/WebGL/WebAudio readback case.
• Every <audio> needs an id. The mixer selects audio[id][src], so an id-less <audio> is never mixed and the render is silent. lint catches it as media_missing_id.
• Audio always lives on a separate <audio> element — even if its source file is the same as a <video>. The <video> is muted; the <audio> carries sound.
• For volume fades and ducking, use the data-automation volume lane; the exact form is in creator-editing-recipes.md. data-volume is the static baseline. A timeline volume tween is ignored when a lane is present.

For media duration: <video> and <audio> can omit data-duration if the media’s intrinsic length is known and you want the full clip. Otherwise provide data-duration explicitly.

Input codecs: render decodes video via FFmpeg (frames are pre-extracted and injected), so HEVC/H.265 assets (8/10-bit) render correctly everywhere; live preview auto-proxies any browser-hostile asset (transcodes and caches an H.264 copy on first use, opt out with --no-proxy or media.autoProxy: false), and lint emits an info-level hevc_preview_codec note naming affected assets.