Filler cleaner
A cleanup that takes every filler word makes the next cut obvious, because nobody speaks that cleanly. Um Out ticks the hesitations that were never part of how you talk, lists the rest, and leaves the decision on those with you.
In the free trial, and on Local ($49 once) and Plus ($7 a month). Needs the local speech model.
The ums are easy to hear and slow to remove, which is why they survive into the upload.
Pure hesitation (“um”, “uh”, “er”, and their stretched forms) is almost never meant, so it is ticked by default.
Hesitation phrases like “I mean” and “you know” are ticked only when the surrounding speech says they are padding.
Context-dependent words (“like”, “actually”, “basically”, “literally”) are listed but never ticked for you. Removing “like” from “it looked like a bug” produces nonsense, and no confidence score is worth that risk.
Confidence here is about the removal, not the transcription. A drawn-out “ummmm” is one of the least cleanly transcribed things in any recording, and for a long time that counted against it: the tool scored it as an uncertain word and declined to cut it, which meant the most obvious filler in the file was the one it skipped.
Pure hesitations are now judged on what they are rather than on how cleanly the model heard them. Everything else still weighs the transcription, because for a real word a poor transcription is a genuine reason not to cut.
Run silence removal first, then fillers. A stumble reads differently once the dead air around it is gone, and some hesitations you would have cut turn out to be fine at the tighter pace.
Picture and dialogue are cut together, so the video track does not fall out of step, and each removal takes the word plus a little either side, capped at half the real gap so words are not welded together.
Yes. The filler word cleaner is one of the three tools in the 7-day trial, with no footage limit and no card required. It needs the local speech model, which downloads the first time you use it.
Yes. Filler removal reads a transcript, which means the speech model has to be installed. The model downloads once and runs on your own machine afterwards, so nothing is uploaded and there is no transcription bill.
Only if you tick them. Anything context-dependent starts unticked, and anything below the confidence threshold is listed rather than selected, so you opt in instead of having a word removed on a guess.
Check the tick boxes in the review list. Words the tool is not sure about, and words like “like” that are real words as often as they are filler, are listed unticked on purpose.
It is on every plan. Local is $49 once for the local editing tools on two computers, Plus is $7 a month and adds hosted AI on top. Filler cleanup is never metered or capped on either.
Seven days, three tools, no card. Run it on something you already shot and judge the result on your own footage.