Source
video-use repository
Read the primary source.
Reference illustration, not a screenshot of this source.
video-use
Claude Code editing skill · Video
A Claude Code skill that edits talking footage from its transcript: picks takes, cuts filler words and dead air, grades, subtitles and renders with ffmpeg.
Use it for
Raw takes of people speaking (a recorded speaker, a demo, an interview) cut into one clean film by asking Claude.
Choose something else for
Not for montages built on music (use family-video), and not for private footage as shipped: the audio goes to ElevenLabs.
Give it
Folder of raw video takes; A plain-English brief
Get back
edit/final.mp4 with optional subtitles, colour grade and overlays; A cut list (edl.json) and a project.md memory
How it is operated
Claude Code skill on the Mac; local transcription tested 15-09-2026, ElevenLabs Scribe optional at $0.22 an hour of audio
Availability
Provider capabilities and historical Studio notes are distinguished below. Current account access, credits and runtime readiness were not tested.
How charging works
ElevenLabs Scribe $0.22 an hour of audio (read 15-09-2026); rendering runs locally.
Where processing happens
Local by default on this Mac (faster-whisper); audio goes to ElevenLabs only when Scribe is chosen.
Language support
Scribe covers more than 90 languages, French and Spanish included.
Cut filmed speech by its words
★★★★☆ 4/5 for this task
The only tool here that cuts at word boundaries, with production checks built into its render (30 ms fades at each cut, subtitles applied last, every cut checked in the rendered file). Tested locally on French speech; subtitles not yet burned.
Decision: Use for talking footage; swap Scribe for the local faster-whisper before any private clip.
Limits and things to check
- With Scribe, every clip's audio is uploaded to ElevenLabs; the local helper keeps it on the Mac but hears one speaker only and misses some fillers and brand names.
- Burned subtitles need ffmpeg with libass; the Homebrew build has none.
- Cuts follow speech; footage without speech gives it little to work with.
Beside family-video, which arranges photos and clips on music, and Video Shotcraft, which films a product's own screens: video-use cuts what people say.
Evidence
Run 15-09-2026 on a 33 s French recording: local transcript, packed phrases and a two-phrase cut rendered; subtitle burn-in not run (ffmpeg without libass)
Catalogue review: 2026-09-15. The evidence note identifies what was verified. This date does not imply a fresh tool test or verification of every linked source.
