AbdalkaderdevAbdalkaderdev/takecut/skills/video-cleanup/SKILL.md

video-cleanup

Skill0 starsChanged 35 days agoMITView on GitHub

Automatically tighten talking-head footage, podcasts, tutorials and screen recordings. Removes silences and dead air, filler words (um, uh, comma-delimited "like", plus German, French, Spanish, Italian, Portuguese, Dutch, Russian, Turkish, Arabic, Japanese and Chinese fillers), stutters and repeated or false-start takes (keeps the last good take), then normalizes loudness to EBU R128. Reports every cut with timestamps and total time saved. Use when the user asks to clean up, tighten, remove ums or pauses, cut dead air, or make a recording sound professional.

In this repository4 other files it gives its agents.
- video-captionsSkill
- video-editSkill
- video-ffmpegSkill
- video-shortsSkill

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.

≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Very specific:
Read cleanup.report.md. Sanity-check retake and stutter rows against transcript.txt.
line 25
Very specific:
Keep more air: --keep-pause 0.25.
line 45

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

SKILL.md

video-cleanup

Detection is a script; the decision stays reviewable: it writes an EDL plus a report, you (and the user) check them, then the video-edit renderer produces the result. Script paths below are relative to this skill's folder (the shared engine is in ../video-edit/scripts/); run them by full path from the user's working directory (don't cd into the skill).

Steps

Transcribe if edits/<name>/transcript.words.json doesn't exist yet: uv run ../video-edit/scripts/transcribe.py INPUT
Detect: uv run scripts/cleanup.py edits/<name>/transcript.words.json -> edits/<name>/cleanup.edl.json and cleanup.report.md, prints counts and before/after.
Read cleanup.report.md. Sanity-check retake and stutter rows against transcript.txt; remove any drop entry from the EDL that cuts real content (each has reason and text).
Preview: uv run ../video-edit/scripts/render.py edits/<name>/cleanup.edl.json --preview
Final: uv run ../video-edit/scripts/render.py edits/<name>/cleanup.edl.json -> cleanup.mp4 and cleanup.words.json (for captions without re-transcribing).
Tell the user: before -> after duration, counts per reason, output path.

What gets cut

reason	detection	what is removed
`silence`	ffmpeg `silencedetect` (`--noise -35dB`, `--min-silence 0.6`)	the silence minus `--keep-pause 0.15` s on each side, so speech keeps a natural breath
`pause`	word gap > ~1.3 s where silencedetect saw no silence (noisy rooms)	the gap minus padding
`filler`	token in the language's filler list; "like" only when comma-delimited	the filler and the pause after it
`retake`	consecutive sentences that are near-duplicates (>= 80% similar) or where one is a prefix of the next	the earlier take; the last one is kept
`stutter`	immediately repeated 1-4 word runs inside a sentence ("I I think", "we should we should")	the first occurrence; "very very", "no no" etc. are kept

Options

Noisy room / music bed: --noise -30dB (higher = more aggressive) or --no-silences.
Keep more air: --keep-pause 0.25; cut only long silences: --min-silence 1.0.
Fillers: --add-fillers "basically,actually" (single words) or replace with --fillers "um,uh"; --no-like.
--no-fillers, --no-repeats, --no-loudnorm, --lufs -14 (social) / -16 (default).
--name tight writes tight.edl.json so several variants can coexist.

Language comes from the transcript. Whisper normally drops "um"/"uh"; transcribe.py prompts it to keep them for en/de/fr/es/it/pt/nl. For other languages fillers may be missing from the transcript; silence and retake removal still work. Per-language lists: references/fillers.md.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free