Repository navigation
v0.1.0 — Transcription, auto-captions, and a real CLI
First tagged release.
Added
bin/transcribe.sh— local Whisper (whisper.cpp by default, or any binary viaVIDEO_WHISPER_BIN) transcript ofingest.sh's extracted audio →transcript.vtt/.txt.review.pynow feeds the transcript (when present) into the Gemini judge prompt for better clip picks.bin/caption.sh—.srtfrom the transcript, plus optional caption burn-in into cut clips (VIDEO_CAPTION_BURN=1).bin/video— one CLI entrypoint wrapping all six stages (pull|ingest|transcribe|review|caption|clip) with--help.
Known gaps (tracked in docs/ROADMAP.md as 🚧)
- Whisper transcription's happy path is implemented but unverified end-to-end (no whisper binary in CI/dev yet).
- Caption burn-in requires an ffmpeg build with
libass; a plainbrew install ffmpegmay not include it.caption.shdetects this and degrades gracefully.
Planned and executed via Hive (/plugin-hive:plan + /plugin-hive:execute) — see .pHive/epics/transcribe-caption-cli/ for the full design discussion and story-level detail.
🤖 Generated with Claude Code