ガイド3 分で読めます
How to remove filler words, pauses and retakes
A practical guide to the rough cut of a long-form talking-head video: what to cut, what to keep, and how to do it in minutes instead of hours.
この記事は英語でお読みいただけます。
The rough cut of a talking-head video is the same job every time: find the long pauses, the ums, the stutters and the sentences you said twice, and cut them without making the video sound chopped. On a 30-minute recording that can take hours by hand. Here is what to cut, what to leave, and how to speed it up.
What a filler word is, and why it matters
A filler word is a sound or word that carries no meaning, like "um", "uh" or "you know". A few make you sound human. Dozens make a video feel slow, and viewers notice the pace long before they notice the words.
What to cut
- Long pauses: silences where you think, look at notes or reset. Cut them down, but not to zero.
- Hard filler words: um, uh, erm. They almost never carry meaning.
- Stutters and false starts: "I, I, I think" or a sentence you abandon after three words.
- Retakes: when you say the same sentence twice, keep the last good take. It is usually the better one, even if the words changed.
- Noises: coughs, throat clearing, a loud breath right before a sentence.
What to keep
- A beat at the end of a sentence. Cutting every pause to nothing makes speech sound breathless. Leave a short gap where a sentence ends.
- Deliberate pauses. A moment of silence before a punchline or a reveal is part of the delivery.
- Soft fillers with meaning. "Like" or "so" are often real words. Only cut them when they really are filler.
- Laughter. It shows the moment worked.
How to make cuts that don't sound chopped
- Cut on the audio, not the transcript. Transcript timestamps are approximate. Place the cut where speech really starts and stops in the waveform.
- Keep a little more after a word than before it. Words trail off; cutting right at the end clips them.
- Watch for hidden pauses. Transcripts often stretch a word over the silence that follows it. Check the waveform, not just the text.
- Hide the jump cuts. Alternating the framing every two or three cuts (for example between wide and a slight zoom) makes them feel like a second camera.
Doing it automatically
Autocut in Lumenwerk does this pass for you. The built-in rules find pauses, filler words and stutters for free on your computer. An optional AI review then reads the transcript like an editor: it catches reworded retakes, abandoned sentences and asides, and leaves the pauses and ums you meant to keep. Every cut lands in the timeline for you to check, and any single cut comes back with one click.
The video file never leaves your computer; only the audio track goes out for the transcript. When the cut is done, render an MP4 in the browser or export the timeline to DaVinci Resolve and keep editing there. To hide the jump cuts, run Auto Zoom on the result.
How much does removing filler words shorten a video?
It depends on the speaker and the recording. Unscripted recordings with many pauses and retakes shrink much more than a tight, scripted take. Measure it on your own footage rather than trusting a general number.
Should I remove every filler word?
No. Remove the ones that slow the video down, keep the ones that are part of how you speak or that carry meaning. A video without a single pause sounds unnatural.
Is it better to cut by hand or with a tool?
A tool does the repetitive pass much faster; a human decides the edge cases. The best workflow is both: let the tool suggest every cut, then review them in the timeline.