MontikScrubbing a waveform for every “um” takes a long time, and a waveform can’t tell you which sound is which word. A transcript with word timings can. Pauses are simpler, because they’re the gaps between words.
This guide uses Montik, the editor we make. The idea carries over to any editor that cuts on a transcript.
Do it by hand
- Upload your video to a Montik project and transcribe it.
- Open the Speech panel. Your transcript sits beside the preview.
- To remove a word, select it and press Delete. The video is cut on the word.
- To remove all filler words and long pauses at once, use the one-click options in the same panel, and set how long a pause has to be before it goes.
Or ask Claude
Connect Claude to Montik, then ask:
Transcribe this, remove the ums and long pauses, and tell me what you cut.
Each change Claude makes is one undo step, so ⌘Z takes back anything you disagree with. Asking it to save a version first lets you compare the two.
Settings that work
Two numbers decide how a pause cut feels: how long a gap has to be before it’s cut, and how much air stays on each side. Montik’s defaults sit between these:
| Video | Cut gaps longer than | Air left on each side |
|---|---|---|
| Punchy social ad | 0.4 s | 0.1 s |
| Default | 0.6 s | 0.15 s |
| Calm explainer or interview | 0.7 s | 0.2 s |
These are starting points. Trust your ear over the numbers.
Be careful with “like”
Montik’s default list covers hesitations such as “um” and “uh”, in English and Slovenian. It leaves out “like”, “basically” and “you know” on purpose, because they’re often real words:
I like this serum.
Read the transcript before adding those. Removing them everywhere can change what someone said.
What to check afterwards
- The first word: a cut near the start can clip it.
- Jump cuts: several cuts close together can look twitchy on a talking head. Laying B-roll over them or punching in slightly hides them.
- Music: cuts ripple through unlocked tracks. Lock the music track first if it should keep playing through.
- Subtitles: add them after the cuts, or generate them again, so they match the final words.
- Meaning and pacing: watch it through once at full speed with the sound on. A natural pause can be doing real work, and taking out every one can make a person sound rushed.
Without Montik
If you’re comfortable in a terminal, the open-source video-use project does similar cuts through Claude Code, on files on your machine. It needs FFmpeg and an ElevenLabs API key for transcription. See our comparison of the approaches.
