jump-cut-editorlisted
Install: claude install-skill yyert123/miragea-skills
# Jump cut editor (silence and filler word remover)
Tightening a raw take by hand takes longer than recording it. This skill finds every dead pause, "um/uh" and stuttered repeat from word-level timing and cuts them with 12 ms audio fades so there are no clicks.
## Steps
1. Check `MIRAGEA_API_KEY` is set (never print it). Free key: https://miragea.ai.
2. Plan first, free of rendering: `python scripts/jumpcut.py talk.mp4 --dry-run` — shows original vs final length and each cut.
3. Tell the user the saved time and cut counts. If it feels choppy, raise `--max-gap 0.6`; keep natural fillers with `--keep-fillers`.
4. Render: `python scripts/jumpcut.py talk.mp4 -o talk-tight.mp4` (audio-only input gives a WAV).
5. Spot-check a frame and the duration with ffprobe before delivering. The transcript is saved next to the media (`talk.transcript.json`); reuse it with `--transcript`.
Next: caption the result with `video-captions`, or cut it into Shorts with `video-to-shorts`. A line that is badly flubbed rather than just padded? Re-record it in the user's cloned voice with the Retake tool.
## What gets cut
| Kind | Rule |
|---|---|
| silence | gaps over `--max-gap` (default 0.45 s) shrink to that length |
| filler | um, uh, er, erm, hmm, mm (not "like", so meaning is never changed) |
| repeat | immediate stutter of a short word ("the the plan"), never across a sentence end |
| tail | dead air after the last word |
## Why not other tools
| | Auto-editor / paid editors | jump-cut-editor