Files
hyperframes/packages/cli/src/commands
Vance IngallsandClaude Opus 5.5 824e4100c6 feat(studio): waveform strip, clip light, peak marks and video beat source (5/8) (#4817)
* feat(studio): draw a waveform strip under video clips that carry sound

An audible video keeps its thumbnails and gains the audio waveform along the bottom
38% of the clip (also when thumbnails are off); muted and silent video are unchanged.
The waveform route already decodes video audio; a test now proves it on an mp4.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* feat(studio): latch a CLIP light on the master meter at the export ceiling

The light turns red when the preview master sample peak reaches -1 dBFS (the export
true-peak ceiling) and stays lit until clicked. Mutation-tested threshold and latch.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(studio): paint the CLIP light from theme colour tokens

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* feat(studio): mark where each clip redlines on its waveform

studio-server computes an absolute per-50ms sample-peak map once per media file
(native channels, cached by size+mtime next to the waveform cache); the timeline
paints red marks where the source at the clip's own volume reaches -1 dBFS and
names the loudest point.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* feat(studio): let a video with sound be the beat source

Studio's isMusicSourceElement accepts an audible, unmuted video for the music track
and the beat fallback (lane zoning unchanged); core findMusicAudioSrc accepts
<video data-has-audio="true"> without muted; the beats CLI message mentions video.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(studio): decode each peak map once and never offer Duck on a carve source

Concurrent peak requests for one file share a single ffmpeg decode; Duck under voice
is withheld on a clip another bed already carves against.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(studio-server): decode the same audio stream the peak map probes

The channel probe read stream a:0 while the decode left stream choice to
ffmpeg, which picks a later stereo stream over a mono first one; every frame
was then read at the wrong width, doubling the bins and taking the other
stream's level. Both commands now select a:0, and the cache key moves to
peaks-v2 so maps built from the wrong stream are rebuilt.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(studio-server): terminate the peak map's ffprobe options before the input

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(studio-server): reject a peak map whose decode exits nonzero

A decode that emits some samples and then fails no longer resolves as a
complete map, so /peaks cannot cache a truncated clipping map.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 04:43:26 +00:00
..