Filler words

The ums, uhs and likes you say while you think of the next bit. They fill a pause but add no meaning, so editors usually cut them out.

Disfluencies, both non-lexical (um, uh, erm) and lexical (like, you know, so), that occupy pauses in spontaneous speech. Editors remove them by cutting the word span from the audio and picture, then smoothing the resulting jump cut with split edits or a reframe.

Fillers come in two families, and they need different handling.

• Sounds: um, uh, erm, hmm. These are never part of the sentence, so cutting them is always safe.
• Words: like, you know, so, right, basically, actually, I mean. Sometimes these are fillers, and sometimes they are doing real work in the sentence.

"It feels like silk" needs its like. "It's, like, really light" doesn't. That difference is why the sounds are easy to remove automatically and the words need a human ear.

A filler is the brain holding its place in the conversation. In real talk, a pause can sound like you've finished, so people fill it to keep the floor. On camera nobody is going to interrupt you, but the habit comes along anyway.

They cluster in predictable places: the start of a sentence, just before a number or a name, and right after you lose your thread. Watching your own footage for those spots tells you where your script is thinnest.

Cut every um and uh. They add nothing, and a viewer hears them as hesitation about the product.

Be choosier with the word fillers. A loose "so" at the start of a casual video can make it sound like you're talking to a friend, which is part of why UGC feels real. Ten of them in thirty seconds sound like nerves. A good rule: keep a filler word only if the sentence sounds stiff without it, and never keep one in your hook, the line that has to land first.

Bang AI handles the two places a filler shows up separately.

• Remove Ums in Auto Edit cuts um, uh and erm out of your lines. It is on when you start.
• Hide ums and uhs, at the end of the Looks tab in Captions, keeps filler words out of the captions, even when they stay in the sound.

So you can leave the audio natural and still have captions that read clean, or cut both. If you ask ChatGPT, Claude or Claude Code to remove all the ums on a video you've already cut, the chat cuts the ones left in the saved video, and fillers removed from a chat don't come back.

Every filler cut out of the middle of a sentence is a cut inside the same shot, a jump cut. Your head is in a slightly different place on each side, and the sound changes at the same instant.

Auto Edit's J and L Cuts option, on when you start, lets your voice run a moment past each cut so the sentence flows across the join. Punch-in Zoom, which you can turn on, zooms in after jump cuts so the change of framing hides the shift. Together they make a line with three ums cut out sound like you said it in one go.

• Know your first five words before you press record. Most fillers come from starting a sentence before you know how it begins.
• Swap the um for silence. A pause is easy to cut; a pause filled with sound is harder.
• Film short lines, one idea each, instead of one long ramble.
• Say a line you stumbled on again from the top. When you film a line more than once, Bang AI keeps every recording, and one reason it picks a recording is Fewest ums. See Swap a clip for choosing another one later.