How to remove filler words from a video (and which ones to keep)

To remove filler words from a video, transcribe it and delete the fillers from the transcript: the matching audio and video go with them. Descript, Adobe Premiere, Riverside and CapCut’s desktop app can find “um” and “uh” automatically and remove them in bulk. On a phone, you can read the transcript and strike the fillers yourself. Either way, do not remove every single one: a video stripped of all fillers often sounds rushed and slightly artificial.

How the main tools do it

Descript

Descript underlines detected filler words such as “um” and “uh” in light blue in the script. To remove them, open the AI Tools panel and choose Remove filler words; a sidebar lists every instance, with timestamps and previews (Descript Help).

You choose what happens to each one:

• Delete removes the word and its audio.
• Delete and replace with gap removes it but keeps the timing.
• Ignore strikes it through in the script and removes the audio, so you keep a record.
• Remove from transcript cleans the text only and leaves the audio.

An option to avoid harsh cuts skips fillers that cannot be removed without clipping the neighbouring words. On current plans, the feature uses AI credits. Descript runs on Mac, Windows and in the browser, not on phones (Descript Help).

Adobe Premiere (desktop)

Premiere’s Text-Based Editing detects filler words in the transcript. Adobe’s announcement describes detection of “uh” and “umm” across all languages supported by Speech to Text, and the transcript panel lets you filter for text, filler words or pauses and delete single instances or all of them at once (Adobe Community). Clips transcribed before the feature existed need to be transcribed again. The same filters find pauses, which Adobe documents separately (Adobe).

Riverside

In Riverside’s editor, open AI tools and turn on Remove filler words. There are three methods (Riverside Support):

• Smart mixes cuts and mutes to keep the flow natural.
• Cut removes the fillers entirely, shortening the recording.
• Mute silences them and keeps the timeline unchanged.

Fillers the detection misses are marked in the transcript so you can remove them by hand.

CapCut desktop

CapCut’s guide to editing video with text describes transcript-based editing in CapCut desktop, with filler words recognised and removed in one click (CapCut).

InShot (phone)

InShot lists an AI Cut feature that removes silence, filler words and breath sounds from footage (App Store).

Tool	Platform	How it finds fillers	Your control
Descript	Desktop, browser	Automatic, underlined in the script	Per word: delete, gap, ignore, text only
Premiere	Desktop	Automatic, filter in the transcript	Single instances or all at once
Riverside	Browser	Automatic, AI tools	Smart, cut or mute; misses marked
CapCut	Desktop	Automatic, transcript	One-click removal
InShot	Phone	Automatic, AI Cut	Applied by the tool
slicecut	iPhone, Android	You read the transcript	Every filler is your call

Why removing all of them sounds wrong

Removing every “um” with one click is satisfying for about a minute. Then the result starts to sound off: sentences slam into each other, the speaker seems to have no time to think, and some cuts land so close to the next word that they click.

Fillers are not all the same. Some stall a sentence and add nothing. Others work like punctuation: a “so” that turns to the next point, an “um” right before an important answer that tells the listener a thought is coming. Linguists Herbert Clark and Jean Fox Tree found that speakers use “uh” and “um” as signals of an upcoming delay, not as random noise (UC Santa Cruz). Our rule of thumb: fillers are punctuation, repetition is noise. The full argument is in filler words are not the enemy.

In practice:

• Remove fillers that stall the start of a sentence, fillers between a false start and its restart, and clusters (“um, so, like”).
• Keep a filler that carries a turn in thought, and most fillers in a relaxed, conversational video.
• Check the cut. If removing a filler makes two words collide, leave it or leave a short gap. Descript’s gap option and Riverside’s mute option exist for exactly this.

How to remove filler words on your phone

slicecut transcribes the take on iPhone or Android, and every “um” and “uh” appears in the text where you said it. There is no automatic removal. You read the transcript and tap the fillers you want gone, which takes less time than it sounds because you are reading, not listening.

1. Import the clip. Only the audio is uploaded for transcription.
2. Read through and tap the fillers that stall. The cut lands right before the next word.
3. Strike the restarts and repeated sentences while you are there; they cost more time than fillers.
4. Turn the pause dial so the gaps are even. This matters: removing fillers leaves irregular gaps, and evening them out is what makes the edit sound natural.
5. Export. Rendering happens on the phone, at source quality.

Tap a struck word and it comes back, so nothing you remove is final until you export.

Remove filler words, or say fewer?

Both. Editing handles the video in front of you. Speaking with fewer fillers makes every future edit shorter. A short training in noticing fillers and pausing instead reduced them sharply in studies of public speakers; the method is in how to stop saying um.

Related: how to remove silence from a video and how to edit a talking-head video.