video-editing

# Video Editing Skill ## Overview This skill implements a **6-layer pipeline** where AI handles judgment tasks (what to keep, what to cut, highlight selection) and FFmpeg/Remotion handle mechanical execution deterministically. | Layer | Name | Mechanism | Primary Tool | |-------|------|-----------|-------------| | 1 | CAPTURE | Inventory source footage | Bash + Glob | | 2 | AI STRUCTURE | Transcript to EDL | LLM judgment | | 3 | FFMPEG CUTS | EDL to segment files | FFmpeg (deterministic) | | 4 | REMOTION COMPOSITION | Segments to TSX composition | Remotion + TypeScript | | 5 | AI GENERATION | Fill gaps with generated assets | ElevenLabs / fal.ai (conditional) | | 6 | FINAL POLISH | Human taste layer | Human + NLE | --- ## Reference Loading Table | Signal | Load These Files | Why | |---|---|---| | `references/preflight.md` | `preflight.md` | Before Phase 1 | | `references/phase-commands.md` | `phase-commands.md` | Each phase | | `references/errors.md` | `errors.md` | Error Handling | | `references/ffmpeg-commands.md` | `ffmpeg-commands.md` | Phase 3, Proxy | | `references/remotion-scaffold.md` | `remotion-scaffold.md` | Phase 4 | | Image-to-video task | `references/image-to-video.md` | Full image-to-video pipeline: validate, prepare, encode, verify | | Image-to-video FFmpeg filters | `references/ffmpeg-filters.md` | FFmpeg filter graphs for audio visualization modes | | Video transcript extraction | `references/video-transcript.md` | yt-dlp subtitle download and VTT cle...