How to remove ums and silences automatically

Removing ums and silences automatically is built into Video by Zaweba. Transcribe the recording in the Transcript tab, then press Remove filler words to take out the ums and ahs and Remove pauses to close up the quiet gaps. Both work on your own PC, and every cut lands on the normal timeline where you can check it and undo it.

It's the kind of job that eats an evening by hand. Spotting every hesitation in a twenty-minute talk means listening to all twenty minutes, often twice, and the two buttons get you most of the way in moments.

Before you press anything

Get the recording onto a timeline and open the Transcript tab. The speech is recognised in pieces of about 30 seconds, all locally, and the words appear as text tied to their place in the edit. The filler and pause tools work from this transcript, so let it finish first.

It also pays to make your content cuts first. If a whole section is going, delete it from the transcript by selecting the words and pressing Delete. There's no point polishing pauses in a passage you're about to throw away.

Running the two clean-ups

Press Remove filler words and the ums, ahs and similar noises are cut out. Press Remove pauses and the silences between phrases go too. They're separate buttons on purpose: some speakers sound better with their pauses left in and only the fillers removed, while a fast explainer might want both.

Then play the result from the top, at normal speed, and listen as a viewer would. Automatic cutting can be too keen. A pause before a punchline or after an important point is doing a job, and taking it out can make a speaker sound rushed. Undo (Ctrl+Z) puts back what you didn't want removed, and because every edit is a snapshot you can step back as far as you need.

Smoothing what's left

Each cut is a tiny jump in sound and, if you're on camera, in picture. A few tools in Video by Zaweba help the result sit nicely. Short sound crossfades soften clicks at the joins; select the edits and use the default sound crossfade with Shift+D. Noise reduction, which learns the noise from each clip's quietest half-second, stops the background hiss stepping up and down between cuts. A voice EQ preset adds clarity to the speech that's left.

On camera, cover the jumpier joins with B-roll on the track above, or use the Smooth jump cut transition, which blends two takes of the same shot with an optical-flow morph.

Where this works best

Talking-head videos, screen recordings with narration, online lessons, interviews and recorded meetings all benefit. It's less suited to drama or music, where silence is part of the performance, and it does nothing for coughs or noises the recogniser doesn't treat as filler; find those in the transcript or on the waveform and cut them yourself.

Once the tidy version plays well, finish as normal: captions from the same transcript, loudness levelled on export, and an MP4 out of Ctrl+E. The ums are the first thing viewers notice, and now they're gone.