edit-videoslisted
Install: claude install-skill akhilmadhav28/edit-videos
# Edit videos — recorded takes → finished explainer
Turn a pile of recorded takes + the script they follow into a produced explainer:
a tight jump-cut edit that removes every dead pause and stumble, a section card on
every topic change, full-screen concept animations that are keyed to the actual
narration, an audio master, soft captions, and a title / description / chapters
that match the channel it's going on.
A full worked example ships in `examples/iit-genai-m1w1-qna/` — 17 screen-recording
takes + a 16-slide script → a 14-minute edit with 16 section cards and 11
speech-synced animations. Its `PLAN.md`, `cuts.json`, filled-in config
(`config-snippets.md`) and all 12 animation compositions are there to copy from.
Blank starting points are in `templates/` (`PLAN.template.md`,
`animation.template.html`).
**Read `reference/gotchas.md` before you start.**
## Prerequisites (one-time)
| Need | Install |
| --- | --- |
| **ffmpeg + ffprobe** ≥ 6 on PATH | `winget install Gyan.FFmpeg` · `brew install ffmpeg` · `apt install ffmpeg` |
| **Node ≥ 20** + `npx` on PATH | nodejs.org / nvm |
| **Python 3.10–3.12** + a venv with `faster-whisper` | `python -m venv .venv` then `.venv/bin/pip install faster-whisper` (Windows: `.venv\Scripts\pip`). CPU int8 runs ~0.2–0.4× realtime — no GPU needed. |
| **HyperFrames** (HTML→MP4) | `npx hyperframes` self-installs; run `npx hyperframes init` once in `animations/hf/`. First render downloads a headless-Chrome shell. Set `HYPERFRAMES_SKIP_SKILLS