Ijharul Islam
To remove filler words (um, uh) from a video, transcribe it, find each filler in the transcript, and delete it so the video cuts that half-second out. Tools like Descript, CapCut, Riverside and QuickVeed find fillers automatically; in any editor you can do it by hand by cutting on the waveform. Then hide the jump cuts with a zoom, b-roll or captions.
This guide shows each method, which fillers to leave in, and how to keep the result sounding natural.
The short answer
Use a transcript-based editor: it marks every um and uh and removes them in one click (Descript, CapCut, Riverside, and QuickVeed's Podcast to Clips for short clips). By hand, find the small waveform bumps between words and cut them. Keep a few natural pauses, and hide the jump cuts with a slight zoom, b-roll or captions.
Key takeaways
- Transcript-based tools find fillers far faster than scrubbing a timeline.
- Remove ums, uhs and false starts; be careful with like, so and you know, which often carry meaning.
- Removing every filler can make speech sound rushed. Leave a breath between thoughts.
- Each removal creates a jump cut on video; zooms, b-roll or captions hide them.
Disclosure: we make QuickVeed and show how to do this in it, along with other ways. Details about other products come from their public pages, checked in October 2026 and linked at the end.
Which filler words to remove, and which to keep
| Word or sound | Usually safe to remove? | Why |
|---|---|---|
| um, uh, er, ah | Yes | Pure hesitation; no meaning |
| False starts (“we, we went”) | Yes | Repeated words add nothing |
| Long pauses | Shorten, don’t remove | Some pause helps the listener; see silence removal |
| like | Check each | Can be a comparison (“like a bank”) or a filler |
| so, well, you know, I mean | Check each | Often connect ideas or soften a claim |
| Laughter, sighs | Usually keep | They carry emotion and make it human |
Automatic tools are reliable on ums and uhs, less so on words like like and so. Review those yourself.
How many fillers are too many? There is no fixed rule. A useful test is to listen once without looking: if you notice the fillers, your viewers will too. In short clips, cut more aggressively; in a full interview, a light touch keeps the conversation natural. Accents and languages differ as well: hesitation sounds vary from language to language, so check that your tool recognises the ones your speakers use.
Method 1: Remove filler words with a transcript-based editor
These tools transcribe the video, tag filler words and delete the matching video when you delete the word.
- Descript: Remove Filler Words and Shorten Word Gaps are on all paid plans and limited on the free one. Good for full episodes.
- CapCut: its filler word remover spots ums and uhs in the audio and lets you preview the edits before export.
- Riverside: the Pro plan and up include removing silences and filler words in its editor.
The general steps are the same everywhere:
- Import the video and let the tool transcribe it.
- Choose the filler words to detect (start with um and uh).
- Review the highlighted words and untick any that should stay.
- Remove them all, then play the result at normal speed.
- Fix any cut that clips the start of the next word by nudging it a few frames.
Method 2: By hand in any editor
In a free editor without filler detection, use the waveform. A filler is a short, low, flat bump between two word shapes.
- Expand the audio track so the waveform is large.
- Play at 1.25× to 1.5× speed and drop a marker at each filler you hear.
- At each marker, zoom in and cut just before and just after the bump.
- Delete the piece and close the gap (a ripple delete).
- Add a 2 to 5 frame audio crossfade if the cut clicks.
Expect about 10 to 20 minutes for every 10 minutes of footage. That is why most people use a tool for long recordings and do it by hand only for short pieces.
A shortcut that helps: cut only the fillers at the start of sentences. Those are the most noticeable, because they delay the point. Fillers in the middle of a sentence are often swallowed by the rhythm of speech and can stay.
Hiding the jump cuts
On camera, removing half a second makes the head jump. A few cuts look fine in short-form video, where viewers expect a fast pace. For longer pieces:
- Punch in: zoom the next shot in 10 to 15% so the cut looks like a camera change.
- Cover with b-roll: lay a screenshot, product shot or slide over the cut while the audio continues.
- Switch angles: in a multi-camera podcast, cut to the other camera.
- Captions: big captions draw the eye away from small jumps.
Audio needs care too. A cut in the middle of room tone can create a click or a sudden change in background noise. Short audio crossfades fix clicks. If the background hum changes at every cut, add a constant bed of room tone under the whole track, or use a noise reduction step before removing fillers.
Doing it in QuickVeed
QuickVeed removes fillers as part of making short clips. In Podcast to Clips, every clip has long pauses shortened and ums and uhs cut automatically, along with the vertical crop, captions and hook title; there is no extra charge for the clean-up.
- Open Podcast to Clips and upload the long video (up to 3 hours). The upload keeps going if you switch to another page.
- Optionally type what the clips should be about, for example “pricing tips” or “hiring mistakes”.
- The AI transcribes the video and returns 3 to 10 moments of 15 to 90 seconds, each with a hook title, a reason it was picked and a score from 0 to 100.
- For each clip, pick the shape (9:16, 1:1 or original), a crop that follows whoever is speaking or a fixed position, and one of 11 caption styles in any language.
- Click Make all clips. Each clip gets the hook title on screen for the first 4 seconds, shorter pauses, ums and uhs removed, and your brand kit’s colour and logo.
- Download them all as a ZIP, or use Post / Schedule to send them to TikTok, YouTube, Instagram, LinkedIn and more (social posting is on the Boost and Scale plans).
Cost: the transcript is 1.5 credits a minute rounded up (45 credits for a 30-minute video, 90 for an hour), and each finished clip is 3 credits with the crop, captions, hook title and clean-up included. A 30-minute episode cut into 10 clips is 45 + 30 = 75 credits, which is exactly the 75 credits of the free 14-day trial.
To be clear about the limit: this is for the clips cut from a long video, not for cleaning up a full-length episode. For full episodes, Descript, Riverside or CapCut are the better fit. To also trim dead air, read how to remove silence from a video.
Prevent fillers at the source
Editing is easier when there is less to remove. Before you record, write the opening line of each segment so you don’t start with um. During recording, pause silently instead of filling the gap; a silent pause is easy to trim. And it is fine to leave some fillers in: a conversation with none can sound scripted.
If you host a show, a few habits help most. Ask one question at a time, so guests don’t restart their answer. Give guests the first question before recording, so their first answer starts cleanly. And when you stumble, stop, pause for two seconds and say the sentence again: the pause makes the bad take easy to find and cut later.
Finally, compare a cleaned clip with the original before you publish a whole batch. If the cleaned version sounds rushed, put back a few of the longer pauses. The aim is a speaker who sounds like themselves on a good day, not a voice with every breath removed.
Frequently asked questions
Can AI remove um and uh from video automatically?
Yes. Descript, CapCut and Riverside detect filler words in the transcript and remove them; QuickVeed does it automatically for clips made in Podcast to Clips.
Is there a free way to remove filler words?
Yes. Cut them by hand in a free editor using the waveform, or try CapCut’s filler word tool. Descript’s free plan includes a limited version.
Should I remove every filler word?
No. Remove ums, uhs and false starts, but keep some pauses and natural words, or the speech sounds rushed and robotic.
How do I hide jump cuts after removing fillers?
Punch in 10-15% on the next shot, cover the cut with b-roll, switch camera angles, or rely on large captions.
Try it with your own video
Make a few clips from your next recording and hear them with the ums and long pauses already cut. Upload one long episode, pick the clips you like, and judge them for yourself. QuickVeed plans start at $18/month, and new accounts get a 14-day trial with 75 credits, no card needed. Sign up or see pricing.
Sources
Competitor details in this post were checked against these public pages in October 2026. Products and prices change, so confirm with each vendor before you decide.
