Removing silences and filler words automatically: which editors do it and how well-documented is it?

Our verdict: Descript for speech clean-up, Filmora for silence only, verify CapCut in-app

If you need both silence and filler-word removal on English speech, Descript documents the most control and the clearest review workflow. If you only need dead air removed from recordings, Filmora documents a Silence Detection tool on a desktop timeline. We could not find CapCut documentation for either feature on its official pages we fetched on 30 October 2026, so do not buy CapCut for this job without seeing the feature in the app first.

FlexClip, outside this comparison's main three, documents a Remove Filler Words option inside its AI text-based editing for uploads up to 40 minutes, but no silence removal on the page we checked.

What each editor documents

How the threshold settings work

Descript's Shorten Word Gaps asks for two things: which gaps qualify, chosen as more than or between a duration range, and the target length to shorten them to, with 200 ms given as an example. It shortens rather than deletes, which keeps a small breath of space between words. You can step through results and shorten one at a time, or use Shorten all.

Filmora's Silence Detection asks for a volume threshold and a minimum duration. The threshold decides how quiet audio must be to count as silence; the minimum duration decides how long it must last. Its guide does not publish default values or units, so test on a short clip first. A threshold set too high treats quiet speech as silence; a minimum duration set too low cuts natural pauses between sentences.

Illustration, not a measurement: a 10-minute narration with 40 pauses averaging 1.5 seconds, each shortened to 0.2 seconds, loses 40 x 1.3 = 52 seconds and ends at about 9 minutes 8 seconds. The same settings on a slow, deliberate tutorial could remove pauses the viewer needed.

Filler-word handling choices in Descript

Descript detects fillers such as um and uh in English transcripts and lists every instance with timestamps and audio preview. For each, you can delete the word and audio, delete and replace with a gap to preserve timing, ignore (strike through and remove the audio while keeping a visible record), or remove from the transcript only and leave the audio. The Avoid harsh cuts toggle skips fillers that cannot be removed without clipping neighbouring words.

By default removal applies to the whole composition. Descript suggests duplicating a section into a new composition, cleaning it there and pasting the result back when you only want part of the file processed. Studio Sound and filler removal are each capped at 30 AI credits per file per application.

Where automatic removal over-cuts

Automatic removal works on sound levels and word timings, not meaning. The common failure modes are predictable, so check for them specifically.

• Breaths removed entirely, making speech sound rushed or robotic.
• Trailing consonants or soft word endings clipped because they fell under the volume threshold.
• Deliberate pauses for emphasis or for a viewer to read the screen removed.
• Jump cuts in on-camera footage where the head position changes between kept sections.
• Background music or room tone cut along with the pause, creating audible gaps.
• Words like so, like or you know removed where they carried meaning.

Review checklist before publishing

Run automatic removal on a copy of the project, then review the copy against the original. A few minutes of review is cheaper than re-uploading a published video.

• Record the duration before and after; investigate if more than you expected was removed.
• Play every edit point at normal speed with headphones.
• For on-camera footage, watch for visible jumps and cover them with b-roll or a zoom if needed.
• In screen tutorials, check that each on-screen action still has time to be seen.
• Restore any filler or pause that carried meaning; Descript's ignore and restore options keep this reversible.
• Re-check captions after cutting, since timing may have shifted.

Sources & verification

Product details and prices can change. Check the linked provider before buying.

1. Descript Help: filler words Accessed 30 Oct 2026
2. Descript Help: Shorten Word Gaps Accessed 30 Oct 2026
3. Descript Help: sound good with AI tools Accessed 30 Oct 2026
4. Descript Help: track and understand your media minutes and AI credits Accessed 30 Oct 2026
5. Descript Help: supported transcription languages Accessed 30 Oct 2026
6. Descript pricing Accessed 30 Oct 2026
7. Filmora user guide: Silence Detection (Windows) Accessed 30 Oct 2026
8. Filmora user guide: AI text-based editing Accessed 30 Oct 2026
9. CapCut Help: CapCut Standard vs Pro Accessed 30 Oct 2026
10. CapCut: auto caption generator Accessed 30 Oct 2026
11. CapCut: video to text Accessed 30 Oct 2026
12. FlexClip Help: AI text-based editing Accessed 30 Oct 2026