[miguel-heygen]

What

A video that carries the muted attribute has its own audio track switched off in the runtime's sync tick, so its picture no longer waits on that track at a cut.

Why

In Safari a muted video is clocked by its audio track. When that track starts late after the cut, the picture stalls and then catches up by rate steering, so the first frames after some cuts arrive 100-250 ms late, and a cue's first frames can be skipped entirely. The sound itself comes from separate audio elements and is not affected.

Related work

Follows the pre-roll revert and the start-latency learning already on main. No issue.

How

dropSilentAudioTrack runs per clip in syncRuntimeMedia: for a <video> whose defaultMuted is true it sets every audioTracks[i].enabled to false. Engines without audioTracks (Chromium) skip it. Videos without muted are untouched; the runtime already keeps defaultMuted videos muted (init.ts mute paths), so nothing audible changes.

Size

A lone fix: nine lines of runtime and its tests; nothing else in flight touches this path, so there is no PR to fold it into.

Test plan

Measured on the iOS 26.5 Simulator by capturing the screen and the sound the app produces, 10-take fixture, 44 scored cues per run, three runs per arm interleaved (load 14-15 at start). Picture minus playhead, + = picture late:

	run 1	run 2	run 3
main (0.9.23) median, within 40 ms	+106, 32%	+124, 5%	+69, 27%
this PR median, within 40 ms	-34, 84%	+17, 70%	0, 70%
flashes seen of 44 (main / PR)	41 / 44	44 / 44	42 / 44

Sound against the playhead stays +18..+24 ms on both.

• Unit tests added/updated: three cases in media.test.ts; fails without the fix: two of the three fail on main (the track stays enabled), the control passes on both.
• Manual testing performed
• Documentation updated (if applicable)
• Comments follow CONTRIBUTING.md "Comments": they say why, not what, and a bug fix says what the code must do and how to reproduce the bug

Small on purpose: a lone urgent playback fix that nothing open can carry.

[github-actions]

Edit accuracy: accurate 267 (base branch 267), smooth 99 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 (8)

• seqplainkeys-none-center-r0-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): fail / pass / pass, passes
• seqplainkeys-none-center-r0-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
• seqplainkeys-none-center-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
• seqplainkeys-none-pct-r0-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): fail / pass / pass, passes
• seqplainkeys-none-pct-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
• seqplainkeys-none-px-r0-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
• seqplainkeys-none-px-r30-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
• seqplainkeys-none-px-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run

Unstable (1)

• seqfastkeys-none-px-r0-root-z100: tracking 0.07, pressJump 0.05, drop 0, reload 0, render 0.02, renderKey -, undo false, teleport true / tracking 0.06, pressJump 0.05, drop 0, reload 0, render 0.02, renderKey -, undo true, teleport true / tracking 0.06, pressJump 0.05, drop 0, reload 0, render 0.02, renderKey -, undo true, teleport true

[heygengenesis]

Requesting changes at 994e2743. CI is red because of this change.

Blocker: typecheck fails. Both the Typecheck and the Test: runtime contract jobs in the CI workflow fail on the new loop:

src/runtime/media.ts(481,9): error TS2532: Object is possibly 'undefined'.
src/runtime/media.ts(481,29): error TS2532: Object is possibly 'undefined'.

Under noUncheckedIndexedAccess, tracks![i] is T | undefined. Iterating the tracks avoids both the index and the non-null assertions:

for (const track of Array.from(tracks ?? [])) if (track.enabled) track.enabled = false;

What I checked and found sound:

• Fails without the fix: I ran media.test.ts with and without the dropSilentAudioTrack(el) call. Without it, "is switched off for a video the author muted" and "is switched off when the browser lists it after the first tick" fail (expected true to be false), and the control "stays on for a video that can sound" passes. With it, all 191 tests in the file pass.

• Video that starts muted and is unmuted later: the helper only ever disables tracks and never turns them back on. That is safe today because no path keeps the same element alive across an unmute:

  • The Studio mute toggle and "has audio" toggle remove the muted attribute through a batch commit, which reloads the preview by default (skipReload only applies to style-only ops).
  • A scene swap keeps a media element only when its authoredShape (all attributes) is unchanged, so removing muted rebuilds the element with fresh, enabled tracks.
  • The runtime, player and slideshow mute paths (el.muted = x || el.defaultMuted, init.ts 4317/6605/6635) never unmute a defaultMuted video, so a disabled track loses nothing audible.
  • Web Audio decodes from the source, not the element's track, and Studio scrub targets the music <audio>.

  It still holds only by those invariants. If a future path removes muted in place without a rebuild, that video would stay silent in Safari. Re-enabling the tracks when defaultMuted is false would close that, but it's optional and not a blocker.

Once the typecheck is fixed and CI is green on the new head, I expect to approve.

[heygengenesis]

Approving at e74ae6d2. The typecheck blocker from my earlier review is fixed.

• Fix: the loop is now for (const track of Array.from(tracks ?? [])) track.enabled = false;. The core tsc --noEmit and the runtime tsconfig typecheck both pass locally, and the CI workflow (Typecheck and runtime contract) is green on this head.
• Tests: all 192 tests in media.test.ts pass, including the new "stays on for a video that is muted by script and unmuted later" case. I already confirmed the "fails without the fix" claim on the previous head: two of the new cases fail without dropSilentAudioTrack(el), and the control passes.
• Muted, then unmuted later: this is still safe for the reasons in my earlier review. An attribute unmute rebuilds the element or reloads the preview, and a script unmute leaves defaultMuted false, so the helper never touches that video.
• CI: CI, regression, CodeQL, Player perf, Player Safari audio, Player Safari perf, preview-regression and PR captures are green. In Windows render verification, the engine-cli-2 lane hit its 35-minute timeout inside the @hyperframes/cli unit tests. Those run in Node and never reach syncRuntimeMedia, and Chromium has no audioTracks, so this change is a no-op there. The core Windows lane passed. I've rerun the failed lane, and that result is still pending.