Learn · Guide 6 of 10
Edit by transcript
Read your video like a document. Delete words to cut them, and clean up pauses, ums and retakes in one review.
Before you start: transcription runs entirely on your Mac. The first time, click Install in the Transcript pane or in HTCUT ▸ Settings ▸ Add-ons (or run htcut models install whisper): it downloads speech recognition (Whisper), the word aligner and ffmpeg in one go, about 2.5 GB. Nothing else to set up. Claude asks before installing it for you.
Transcribe
Open the Transcript pane ⌘6 and click Transcribe. Opening the pane never starts it on its own. Each source file is transcribed once and cached. Tag your voice row Dialogue so HTCUT knows which row is speech.
Read along
The transcript follows playback. Click a word to jump there; ⇧-click to select a run of words. Pauses show their length.
Cut words
Select words and press ⌫. The speech and any clips synced to it close up; music on other rows simply shortens. Each cut lands on the quietest 10 ms near the word edge, so you don’t hear a click or half a syllable.
Clean up in one pass
Click Clean up…. It finds long pauses (shortened, not removed), fillers like “um”, “uh” and “hmm” (not “like” or “you know”), doubled words and retakes, keeping your last take. Untick anything you want to keep, click a row to hear it, then Remove.
$ htcut transcript clean --dry-run
$ htcut transcript clean --pause 0.6 --keep 0.25 --no-retakes
$ htcut transcript cut --text "so, so"Catch what slipped through
htcut stumbles looks for doubled words, restarts and fillers that are still in the cut, across every voice row, using word-for-word transcripts. htcut check warns when a clip edge clips a word, and htcut fix-speech extends those clips to the end of the speech.
$ htcut stumbles --transcribe
$ htcut stumbles --fix --dry-run
$ htcut check
$ htcut fix-speech --dry-runCommand only: stumbles and fix-speech aren’t in the app’s menus yet. Clipped words also show up as a blocker in the Pre-publish Checklist.
Animated captions
htcut captions transcribes your voice row and writes word-group captions with the spoken word highlighted, onto their own row. Running it again replaces that row, so re-run it after you re-record.
$ htcut captions --track Voice --into Captions --words 3 --accent "#FFD60A"