Cutting every pause and filler word by hand takes hours. Here it is two scans: each one proposes cuts, you untick anything you want to keep, and one click applies the rest with everything downstream rippling into place.
Select the talking clip Click the video clip, or the voiceover on an audio lane. Both tools work on either, and both respect your trims and splits: only what is on the timeline is analysed.
Find silences In the Audio tab, Remove silences scans for quiet stretches, free and with no tokens. Each proposed cut shows its time and length, so you can untick any pause you want to keep for pacing.
Cut them One click removes every ticked silence. Cuts ripple across all lanes so linked audio and music stay in sync, and a single Cmd/Ctrl+Z restores everything.
Find filler words Remove filler words transcribes the speech at word level, with the token price quoted first, and lists every "um", "uh" and similar sound with its timestamp.
Approve and cut The same checklist flow applies: untick any you want to keep, because sometimes an "um" is characterful, and cut the rest. Each cut leaves breathing room so words are never clipped.
Tips
Run silences first, then filler words: the filler scan is billed on the clip's length, so a tightened clip costs less.
The word list is kept narrow (vocal sounds only), so it never proposes cutting a word like "like".
Back-to-back fillers ("um, uh") merge into one clean cut.
Both scans work on split and merged clips, because the played material is what gets analysed.