Remove Silences From Video Automatically, Keep the Breaths

To remove silences from a video automatically, an editor finds the gaps between words in a transcript with word timings and cuts them out. It leaves short pauses where a breath sounds natural. The cut points come from the words, not from a volume threshold, so a soft word never gets clipped. Then false starts and repeated takes go too, and the jump cuts get hidden with small zooms or b-roll.

Why does a volume threshold cut badly?

The classic silence remover listens for audio below a level and deletes it. It works on clean studio audio and fails everywhere else:

• Soft words disappear. The last syllable of a sentence often falls under the threshold, so it gets cut mid-word.
• Room noise isn’t silence. A fan, traffic or music keeps the level up, and nothing gets removed.
• Every gap is treated the same. A dramatic pause before a punchline goes out with the “um”.

Cutting on words fixes all three. A transcript with the start and end of every word tells you exactly where speech stops and starts, whatever the room sounds like.

How long should the pauses you keep be?

Gap between words	What it usually is	What to do
Under a quarter second	Normal speech rhythm	Keep it
A quarter to half a second	A breath, a comma	Keep it in a calm video; trim it in a fast one
Half a second to a second	A full stop, a thought	Trim it to a short pause
Over a second	Thinking, looking at notes	Cut it
A repeated sentence	A retake	Keep the last take, cut the others

These are starting points, not rules. A fast reel can run tighter; a calm explainer needs the breaths or it sounds rushed. Our post on reel length covers pacing by format.

What about “um”, false starts and retakes?

Silences are the easy part. The bigger win is cutting what you didn’t mean to keep:

• Fillers. “Um”, “uh”, “like” at the start of a sentence. Cut them when they stand alone; keep them when cutting would change the meaning.
• False starts. “So the thing is… so what you want to do is”. Keep the second, complete version.
• Retakes. When you said a line twice, keep the better take, usually the last one.

This needs reading, not measuring, which is why a transcript beats a waveform.

How do you hide the jump cuts?

A talking head with twenty cuts can feel jittery. Four ways to smooth it:

1. Alternate the framing. Every few cuts, punch in a little (a slow zoom on the next segment) so the cut reads as a camera change.
2. Cover cuts with b-roll. Put a shot of what you’re talking about over the cut. Our stock footage guide covers where to find it.
3. Cut on the beat of the sentence, not in the middle of a gesture.
4. Keep captions running. Word by word captions carry the eye across the cut. See word by word captions.

How does Cut Express remove silences?

Cut Express is a video editor you drive from Claude. You upload the recording in the chat and say what you want.

1. Transcript with word timings. Claude asks Cut Express to transcribe the clip (0.1 credit per minute of audio, once per file). It gets every word with its start and end.
2. The cut list. Claude reads the transcript and keeps the lines you want. Long gaps, false starts and retakes go. Each kept stretch becomes a scene with exact in and out points.
3. The polish. Scenes alternate with small zoom-ins, b-roll from Pexels covers some cuts, and captions follow the words through every cut and speed change.
4. Notes. “Too tight, keep more breathing room” or “cut the part about pricing” renders the next version.

You don’t set thresholds. You say how you want it to feel: “tight like a TikTok” or “calm, keep the pauses”. Every version stays in your account, so you can compare a tight cut with a calmer one before you post.

What should you ask Claude for?

“This is a 6-minute recording of me. Remove the long pauses and the false starts, keep the best take of every line, cut it to under 90 seconds, zoom in a little every few cuts, and add pop captions.”

For a podcast, our guide to podcast clips shows how to pull the best minute from an hour.

FAQ

Does it work in languages other than English? Transcripts detect the language, and the word timings work the same way.

Will it cut my words in half? Cuts land on word boundaries from the transcript, so a word is either in or out. You can ask for a few frames of padding at each cut.

Can I keep some silences on purpose? Yes. Say which ones: “keep the pause before the punchline”.

What does it cost? The transcript spends credits once per file; renders spend render minutes equal to the video’s length. See pricing.

More answers in the FAQ on the home page.

Tighten your next talking head with Cut Express

Cut Express is the video editor you drive from Claude: upload the recording, say how tight, get the MP4. See the plans.