Cut video stutters and flubs
/cut-mistakes
/cut-mistakes is a Claude Code skill in the Video & Motion section. It is maintained by nateherkai and published under the MIT license. Finds stutters, repeated words, false starts and retakes in a talking-head video and cuts only the ones you approve.
Author's description
Agent 2 of the video editing pipeline. Finds and removes spoken mistakes — stutters, repeated words, false starts, and retakes (re-recorded lines) — from a talking-head recording. Use after cut-silences, when asked to cut mistakes, remove stutters / repeats / filler restarts, clean up flubs, or keep the best take. Works review-gated: it proposes every cut with context and a reason for approval, then renders only the approved cuts via ffmpeg. Requires a word-level transcript.
Use it when
- You recorded a talking-head video and want to clean up verbal slips
- You said a line several times and want to keep the good take
- You have already trimmed the silences and need the second editing pass
Not for
- Removing silences and pauses: that is what cut-silences does
- Deciding on its own: every cut goes through your review because emphatic repetition can look like a stutter
What you get
A list of cut candidates with context and reason, a file of approved cuts, a re-timed transcript and, once applied, the trimmed video rendered with ffmpeg.
How to ask for it
Remove the stutters and repeated words from my latest talking-head recordingReview this recording and propose cuts for the false starts for me to approveClean up the flubs in this take and keep only the best version of each line
Install
curl -fsSL https://raw.githubusercontent.com/sgomez-dev/claude-skills/main/install.sh | bashAfter installing with the script, type /cut-mistakes. Using Cursor, Windsurf or Codex? Platform guides
Questions about this skill
- Does it cut the mistakes automatically?
- No. It proposes each cut with its context and reason, and only renders the ones you approve, because deciding what counts as a mistake takes judgment.
- What does it need to run?
- A word-level transcript, normally the one cut-silences produces, plus the video if you want the trimmed result rendered with ffmpeg.
- What if my video is already clean?
- It may find no real cuts, and that is a valid outcome: it will not trim natural speech just to hit a quota.
Demo coming soon