v0.2.2

[v0.2.2] — 2026-07-28

Features

• bench: stream-scale bench for the #566 gate (folds in #475's measurement) (#787) (0c8ed52), closes #458 #754
• export: optional audio-track choice on ExportVideo/ExportAudio — single stream, no mixdown (#561) (#769) (474d7a0), closes #560
• labels: range labels — annotation kind 'label' + transcript UI (#543) (#782) (26e686a), closes #548 #563
• media+ui: audible track audition — sampled windows, explicit segmented UX (#751) (#759) (1180c1e), closes #754 #754 #564 #560 #754 #754 756/#757 #560
• media+ui: voice-track suggestion — burst/quiet scoring, preselect-only (#750) (#760) (857ae23), closes #751
• media: per-source voice-track selection for transcription/waveform (#560) (#614) (1f7633a), closes 564/#565 #317 #561 #748 #447 #633 #748
• media: reaction-marker detection — rolling-p80 baseline over the voice track (#563) (#766) (fd78ce8), closes #560 564/#755 #562 #769 #561
• media: streaming waveform decode — O(buckets) memory, no whole-file PCM (#564) (#615) (df36531), closes #560 #560 #566
• media: windowed waveform peaks — zoom-aware fetch, O(window) decode (#754) (#758) (4b0d6de), closes #564 #560 756/#757
• reactions: accept reaction candidates into range labels (#563) (#791) (deba055), closes #543 #548
• ui: auto-transcribe the detected voice track, and say so (#763) (#765) (f2b1f46), closes #750 #561 #749
• ui: export-panel audio-track selector behind VITE_SCENARISTO_EXPORT_AUDIO_TRACK (#767, #768) (#780) (0432b72), closes #560 #561 #769
• ui: reactions review card behind VITE_SCENARISTO_REACTIONS (#563) (#777) (ac73cbb)
• ui: the popover asks one question again — voice track gets its own home (#749) (#764) (e958c9b), closes 750/#751 #748
• ui: the strip says it is reading audio instead of drawing silence (#796) (#797) (1afed66), closes #566 #793

Bug Fixes

• ci: switch Windows code signing from PFX+signtool to SSL.com eSigner (#786) (c3aac84)
• edit: bulk filler strike rides one core compound, never a covering band (#752) (#756) (c2ff45b), closes #633
• media: clamp libav's log level — the sampled-window seams flooded stderr (#761) (907765d), closes #759 #751 #750 #754
• media: decoded pixels honor the display matrix (#778) (#785) (056f62c), closes #503 #665
• media: distrust an implausibly sparse container index (#788) (#790) (e0907a8), closes #458 #466
• media: probe reports display dimensions for rotated media (#665) (#781) (c652520), closes #778
• media: reaction detection tuned on real footage — baseline floor + chosen-length clips (#562/#563) (#776) (1ed1be8), closes 515/#281
• media: stream the loudness measurement — no whole-timeline PCM (#755) (#757) (ed3daa2), closes #474
• observe: classify exportVideo/measureLoudness as heavy IPC ops (#745) (5d37bcf), closes #742 #743 #744
• telemetry: tag each Sentry release with the shipped tag, not the frozen crate version (#775) (2167624)
• ui: an empty project looks empty — real waveform, no stale card (#772) (c1d19d0), closes #754 #241
• ui: clicking a label's time scrolls the transcript to it (#543) (#795) (258260b), closes #777
• ui: drop the rail's Change button — it duplicated Re-transcribe (#749) (#774) (e050a30)
• ui: every modal answers Escape and Enter (#560 field report) (#771) (a83d704)
• ui: one reference-counted core://event subscription (DESKTOP-4) (#779) (9df1759), closes #770 #762 #753