[vanceingalls]

What

Carries capture motion blur through renders and Studio, and tightens which elements the scan blurs.

• Engine: installPageMotionBlur sets PAGE_MOTION_BLUR_OFF_GLOBAL before any page script runs, when the engine's own sub-frame blur is on (--motion-blur) or when the render must match a Studio preview that could not blur, and tells the page the render's device scale. CaptureOptions.pageMotionBlur carries the request.
• Producer: a composition whose root (the one the runtime would pick: data-root="true" first) sets data-motion-blur="on" gets a captureMotionBlur render-mode hint and goes through screenshot capture, because html-in-canvas needs a real paint; it is not pinned to one worker. A script-set attribute is not seen: it has to be in the HTML. The probe session carries the request's pageMotionBlur like every other session, so "off" survives a render with no data-duration.
• Studio and Studio server: each render request carries previewMotionBlur, true only if Studio's own browser has drawElementImage and requestPaint. The render route turns previewMotionBlur: false into pageMotionBlur: "off" once, so the Studio dev server and the CLI server both receive the final value.
• Mover scan: a large element that is see-through while it moves (a translucent sweep, a light leak, a large image or canvas with alpha) is no mover, because its crop would carry the scene beneath it along its path; its fast children still blur. An opaque large photo still blurs; <video> counts as opaque. The scan also skips elements moved by a 3D rotate axis or a z translate.
• CLI snapshot: injects video frames before the seek settles, matching the engine's order.

Why

Part 4 of 9 in the capture motion blur stack: #5875 → #5967 → #5968 → #5969 → #5970 → #5971 → #5990 → #5991 → #5992. It builds on #5968; this is where a render can use the runtime.

Stack goal: opt-in capture motion blur for compositions, in Studio preview and in renders, drawn with Chrome's html-in-canvas. Nothing blurs without data-motion-blur="on" on the root.

Notes

• WYSIWYG: when Studio sends previewMotionBlur: false the render has no blur, even for opted-in compositions.
• Engine sub-frame blur (--motion-blur) and page blur never both run.
• Cost: about 1.7x median render time for a blurred composition. k3-promo, 16.5 s, 1080p30: no blur 13 s, capture motion blur 28 s, engine --motion-blur 27 min. Compositions that are blocked or static cost nothing extra.
• Output can differ across worker counts, at about the size of the chunk-split noise a render without blur already has; making every worker's scan find the same movers is feat(core,producer): decide blur defects once per timeline, off on routes that cannot host it #5990.
• <video> in Studio preview: a video drawn through the capture is about 2–3% darker than normal playback, visible as a small brightness change on play/pause. Renders are unaffected (video frames are injected as images). Accepted.
• Not detected as see-through movers, so they are treated as opaque: a CSS background-image, mask-image or clip-path that lets the scene show through; an element that is see-through over less than half of the scene is not recognised as a sweep. A large image or canvas whose pixels cannot be read (cross-origin) counts as see-through and is not blurred. All documented, not fixed.
• Limits that apply to what a render now shows: trails are drawn over foreground elements above a mover; a transparent mover carries a halo of its background inside its box; trails leak past overflow:hidden; a 2D flip through scaleX = 0 blends as rotation; sub-pixel left steps land in whole pixels; a fast child inside a parent that moved earlier is not blurred; SVG elements are not scanned.
• Not in this PR: blur does not yet check for the things html-in-canvas captures wrongly (feat(core): block capture motion blur on html-in-canvas defects #5971); the engine side that tracks WebGL canvases before page scripts is feat(core,engine): track accelerated canvases before any page script runs #5970.

Before

k3-promo at 9.1 s, no motion blur:

After

Same frame with capture motion blur (360° shutter): the pins, the text and the box smear along their paths.

Testing

• Engine pageMotionBlur.test.ts: the off switch for engine blur and for an unsupported preview, device scale.
• Producer htmlCompiler.test.ts: the render-mode hint on the runtime's root, not applied when the preview could not blur; probeStage.test.ts: the probe session carries pageMotionBlur; renderOrchestrator.test.ts and request validation: pageMotionBlur is passed through and only "off" is accepted.
• Studio-server render.test.ts and CLI studioServer.renderBrowser.test.ts / Studio vite.adapter.render.test.ts: the adapters receive pageMotionBlur: "off" for previewMotionBlur: false; Studio useRenderQueueMotionBlur.test.tsx: the request body carries it.
• Core discover.test.ts: translucent, fading, colour and paint-less sweeps are no mover, opaque ones stay, large images and canvases with and without alpha, 3D individual transforms; CLI snapshot.test.ts and captureCompositionFrame.test.ts: the order of video injection and seek.
• bun run build and tsc --noEmit in core, engine, producer, cli, studio-server and studio pass; comment ratchet against feat(core): harden the capture motion blur runtime lifecycle #5968's branch passes; no package size budget changes.

🤖 Generated with Claude Code

[miguel-heygen]

Delta review at full headfcf0512a4427331842036cdc7f73da6eb142ccd0, relative to68737fe2.

Audited: htmlInCanvas.ts form detection/marking, package subpath exports, defect predicates/version bounds, blur wrap/unwrap marking, VFX capture-form adoption and Studio capability consumer. Detection prefers layoutSubtree before drawable/updateElementGeometry, and nested-canvas checks include both forms. Defect version bounds match the author's measured claims; the next PR intentionally makes them conservative across versions.
Trusting: the supplied real-Chrome defect measurements and private test audit; no fresh browser comparison by Home.

The inherited #5875 pose and #5967 timeout findings remain; this delta approval does not clear the complete stack. Saved/fresh prior human reviews and inline comments were empty. Exact-head hosted checks pass/skip; the private focused core run passed124 authored tests. Inspected child regression logs execute3 visual/audio cases, not every new unit test. Browser-version screenshots, HDR output, worker parity, colour and performance measurements were not independently rerun. No provider call, merge or release.

— Magi (Home)
Verdict: APPROVE
Reasoning: No additional blocking defect found in this change relative to its stack parent; the foundation fixes remain required.

[miguel-heygen]

Reviewed exact head 774f8abccd4d0290d2c28257d259ad435a15145d relative to 3a35ca04264c42a87c610b9ecfa15620b3629569.

Audited preview availability/explicit false through Studio, server, CLI adapter, producer probe/capture and engine pre-navigation installation; compiler root/opt-in hint and video injection before paint settlement. False remains off end-to-end; absence/true preserves root opt-in. Read changed integration sites in larger producer/engine files, not every unrelated existing line. Author browser/colour/performance measurements were not rerun.

Independent revision-bound review of the nine-PR composition. Earlier approvals/requests covered older heads. Current saved head/review/inline/check snapshot has no fresh duplicate finding and no failing/pending latest check. Stacked broad-suite skips are not executed tests.

Audited: Studio capability/off choice through server, CLI adapter, producer probe/build/capture and engine pre-navigation installer; compiler opt-in/root choice and video injection before paint settle.
Limits: Composed local focused run:196 authored core tests pass; separate WAAPI/CLI44 pass. Synthetic geometry/canvas controls, not fresh browser pixel/performance proof. No app/model/provider call, source fix, merge or release.

— Magi / desktop-review

Verdict: APPROVE
Reasoning: No additional blocker in this own delta. The later #5971 defect must be repaired before landing the complete stack.

The base branch was changed.

[github-actions]

Edit accuracy: accurate 271 (base branch 271), smooth 140 of those

The gate passes.
Smoothness is reported in the artifact, not gated. A case fails only if it fails 2 of 3 runs.

Quarantined, measured but not gated (0)

Unstable (1)

• seqplainkeys-none-pct-r0-root-z100: tracking 0.06, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo false, teleport true / tracking 0.06, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo true, teleport true / tracking 0.06, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo true, teleport true