How to remove filler words from a video, and what to keep
The quickest way to remove filler words is to edit from a transcript. Find the “um” and “uh” in the text, cut those spots, and smooth the audio at each cut. Don’t cut all of them, though. A talk with every hesitation removed tends to sound hurried and a little strange.
What counts as a filler word
Some are easy calls. Others depend on the sentence.
| Type | Examples | Usual call |
|---|---|---|
| Pure fillers | um, uh, er | Cut |
| Verbal tics | you know, I mean | Cut most of them |
| Hedges | like, kind of, sort of, basically | Judge each one |
| Openers | so, well, okay so | Often keep |
| False starts | “we went to the, the store” | Cut the first try, keep the clean one |
| Dead air | long silences | Shorten, don’t erase |
“Like” is the tricky one. In “it was, like, huge” it adds nothing, so cut it. In “I like this one” it is the sentence. A method that works from a word list alone has to guess which is which, so listen to what it did with every “like”.
Three ways to do it
| Method | Good for | Watch out for |
|---|---|---|
| By hand in a timeline editor | Short videos, full control | You scrub, cut and close the gap for each one, which gets slow on a long recording |
| Editing from a transcript | Long videos of people talking | The cuts follow the text, so the audio can still click at a cut |
| Automatic removal | Quick cleanup of a long recording | It makes judgment calls for you, so you need to listen to the result |
Pick by length. For a short clip, doing it by hand is fine and teaches you what sounds natural. For a long recording, you’ll want the transcript.
Cut it so it doesn’t sound cut
A cut that you can hear is worse than the filler you removed.
- Add a very short audio crossfade at each cut. Without one, a hard cut can leave a tiny click.
- Leave a breath. Cutting right up to the next word makes speech sound clipped. Keep a little room.
- Cover the jump on camera. Taking a word out of a talking-head shot makes the picture jump. Hide it with a punch-in, a cutaway or b-roll, or lean into the jump cut if that’s your style.
- Keep the pause before an important line. It’s doing work.
- Check at normal speed. A waveform won’t tell you whether it sounds natural. Your ears will.
A worked example
Here’s the sample line from our homepage demo. It’s a made-up sentence, not a real recording:
So, um, the real trick is, like, you just, uh, hit record and, you know, keep going.
The “um” and “uh” go without a second thought. The “like” is a hedge that adds nothing here, so it goes too. “You know” is a tic, and it goes. “So,” stays, because it tells the listener a new thought is starting.
That leaves: “So, the real trick is, you just hit record and keep going.”
Cut the “so” as well and you get a clean sentence that no longer sounds like the person who said it. That’s the usual cost of cutting everything.
Silence and dead air
Shorten the silences that make a viewer check whether the video froze. Leave the short ones alone. A beat after a joke, or before a big claim, is the speaker’s timing, and removing it flattens the delivery.
When to leave the fillers in
- Interviews where the hesitation is the point, such as a hard answer.
- Comedy, where the pause is part of the joke.
- Recordings that have to stay as they were, such as an archive.
Where ClipBomb fits
When ClipBomb exports a whole episode, it can remove filler words and silences, as well as cut short clips. Transcription runs on your Mac, so the recording doesn’t have to be uploaded for this step. The first post on this blog explains what does leave your machine. To hear when it’s ready, join the waitlist below.
An automatic cut still needs a listen before you publish, and that goes for ours too.
Join the waitlist — one email when ClipBomb is ready.