sqwartHome

Remove Filler Words From Video

Tighten a talking video by cutting filler words and long pauses, all on your own device.

  • Free
  • No upload
  • No watermark
  • No sign-up
  • AI runs on your device
How it works

How to remove Filler Words From Video

  1. 1

    Choose a video or audio file of someone talking.

  2. 2

    Press Find the words; Whisper writes a word-timed transcript.

  3. 3

    Review the fillers and pauses it found and untick any to keep.

  4. 4

    Press Cut them and download the tighter file.

In detail

How it finds each word

Whisper, running in your browser, writes a transcript with a start and end time for every single word. The tool looks through it for the words on your list, shows each one with the words around it, and lets you untick any you want to keep. Only then does ffmpeg, also running in your browser, make the new file. Nothing is uploaded at any step, which matters for unreleased videos.

The models are the same Whisper sizes as our transcription tools, in versions that also report word timings: Fast (64 MB), Balanced (136 MB) and Accurate (250 MB, best for Hindi and noisy audio), downloaded once and kept by your browser. Word timings are usually within a tenth of a second; a small margin is added around each word so its edges are covered.

What it cuts, and an honest caveat

Two things make talking videos drag: filler words and dead air. The tool cuts every word or phrase on the filler list (basically, actually, literally, you know, I mean, and Hindi fillers such as matlab and yaani, in Roman and Devanagari) and shortens every pause longer than the limit you set to 0.4 seconds, so the speech flows without sounding clipped. The list is yours to edit, and every cut is shown in context before anything changes.

The caveat: Whisper was trained on tidy transcripts, so it usually leaves out "um" and "uh" rather than writing them, and a sound that isn't in the transcript can't be cut. They're on the list and are removed when Whisper does write them, and the pauses around them are caught by pause shortening, but expect some to remain. Word fillers and pauses are handled reliably.

Speed and output

  • Audio files are cut quickly and saved as MP3, WAV or M4A, matching the original where possible.
  • Videos must be re-encoded after cutting, which in a browser takes roughly as long as the video plays, or longer for 1080p and 4K. Keep the tab open.
  • The result is a single file with jump cuts. Very frequent cuts can look abrupt on camera; raise the pause limit for a gentler edit.
FAQ

Questions, answered.

Does it remove um and uh?

When Whisper writes them, yes. Often it doesn't write them at all, so some will stay; long pauses around them are still shortened.

Which filler words are on the list?

Um, uh, er, hmm, like, you know, I mean, basically, actually, literally, kind of, sort of, so yeah, and Hindi matlab and yaani. Edit the list freely; 'like' and 'actually' are often meaningful, so review those.

Is it free and private?

Yes. Transcription and editing both run in your browser; your video is never uploaded.

Will it work on Hindi videos?

Yes. Choose Hindi and the Accurate model; Hindi fillers in Devanagari are on the list.

Why is the video slow to export?

Cutting means re-encoding the video, and the browser's version of ffmpeg is slower than a desktop editor. Audio-only files are fast.

Keep going

More free video tools