Remove filler words from a transcript

The filler list this counter uses

The sound fillers and the stutter rule are the same ones UGCLive's AI edit applies to a recording (its rules file, read on 2026-10-13). AI edit cuts these automatically because they never carry meaning. Word fillers are different: AI edit leaves them in, because “like”, “so” and “actually” are often real words, so here they're marked for you to decide.

Simple context rules skip the obvious real uses: “like” only counts when it's set off by commas (“I like it” doesn't), “so” and “well” only at the start of a sentence, and “you know” not after words like “do” or “if”. Text can't show tone, so read the highlights before you trust the count.

How to remove filler words from a video

In a 30-second ad every half second counts, and nobody is going to interrupt you, so the gaps fillers hold open have no job to do. Five ways to get them out, from cheapest to latest:

1. Before you film: script the opening

   Most fillers land where you're thinking of the next line. A written hook and call to action, even if you riff the middle, takes away the worst gaps.

2. While you film: pause instead

   Silence is easy to cut; an “um” over a word is not. When you lose your place, stop, breathe and start the sentence again.

3. Read from a teleprompter

   Reading at a steady pace removes most of the thinking gaps. Keep the speed near your natural one (about 150 words per minute) so it doesn't sound read.

4. After you film: cut the sounds

   Um, uh and erm can be cut without changing what you said. Cut on the gap either side so the jump isn't heard, and cover visible jumps with B-roll.

5. Then judge the word fillers

   Like, so and actually are often doing a real job. Read the transcript and only cut the ones that add nothing; the rest are part of how you talk.

Read more in our guides to filler words and how to cut them and using B-roll to hide jump cuts, or rehearse with the teleprompter and check your timing with the script template's timer.

What the percentage tells you

The percentage is fillers divided by all words, and the per-minute figure assumes you speak at 150 words per minute. There's no official “good” number: compare your own takes over time, and pay most attention to sound fillers and stutters in the first three seconds, where viewers decide whether to keep watching.

In the UGCLive app, AI edit does this on the video itself: it transcribes the take, cuts the sound fillers, stutters, retakes and long pauses, and you choose how hard pauses are cut.