How do I remove filler words like um and uh from a video automatically?

The fastest way: ask for it in your brief

In Shorty, filler removal is part of making the edit, not a pass you run first. Add your footage, describe the video you want, and include a line such as “remove the ums, uhs and false starts.” The director transcribes the speech word for word, decides which words are fillers in context, and cuts them from both the picture and the sound.

Because it happens inside the edit, everything timed to the speech follows the clean take. Captions, sound cues and the edit’s length are rebuilt from the cut transcript, so nothing drifts after the cuts. You can also ask for it later, in a revision of an edit you already made.

Doing it in Descript, CapCut or Premiere Pro

These editors have a dedicated filler-word tool. Each works on a transcript: you generate it, review what was detected, and delete it before or during the rest of the edit.

Sources: Descript: filler words · CapCut: filler word removal · Premiere Pro: detect and delete pauses and filler words. Competitor features checked 20 October 2026 in each tool’s own documentation. Plans and features change, so check the current page before you choose.

What to remove, and what to keep

Remove hesitations (um, uh, er), stutters and repeated words (“the, the”), false starts where a sentence is abandoned and restarted, and verbal tics that carry no meaning. Keep a word when it does work in the sentence: “like” in “businesses like Stripe” means “such as,” and a repeated word can be deliberate emphasis.

Don’t remove every pause. A breath between thoughts sounds natural; a talk with every gap closed sounds rushed.

Check the result

1. Listen at each cut for clipped words or a click.
2. Watch the speaker at each cut. Cover a visible jump with B-roll or a caption.
3. Play the captions against the speech after the cuts.
4. Read the transcript of the finished video to confirm no meaning was lost.

Why automatic tools miss some ums

A tool can only cut a filler that appears in its transcript. Many speech recognizers are trained to produce clean, readable text and leave hesitations out. When we ran a widely used transcription model on six interview excerpts with audible hesitations, its transcripts contained no “uh” or “um” at all. Shorty uses a verbatim recognizer that writes them down, so they can be found and cut.

See it on four real talks

We ran four public talks through the same tool the director uses inside an edit. Across them, 107 words came out in 80 cuts, including every “um” and “uh” in the transcripts, and every kept word survived in order. The launch post plays each one before and after, with every removed word struck through.

Sources: Shorty News: four talks before and after. Measured on excerpts of public YouTube talks on 20 October 2026. Results depend on the recording.