# How to remove filler words from a video

Canonical URL: https://getvidiyo.app/blog/remove-filler-words-from-video
Last updated: 2026-10-08
Author: Ajay Pawriya (https://getvidiyo.app/about)
Published: 2026-10-08
Category: Talking head

> TL;DR: Transcribe the video with word timings, find the fillers (um, uh, false starts, doubled words) in the transcript, and cut each one on its word boundaries. In Vidiyo, type "cut the ums" and the assistant deletes them by word. By hand, search the transcript or listen through and split around each filler.

Filler words are the "um", "uh", "you know" and half-started sentences that fill thinking time when you talk. On camera they make you sound less sure than you are, and they add up. The fix is simple in principle: find each one and cut it out without clipping the words around it. Here's how, with AI and by hand.

## What counts as a filler word?

A filler word is a sound or word that carries no meaning in its sentence: "um", "uh", "er", a doubled "I, I", "so" at the start of every sentence, or a false start you abandoned. The same word can be a filler in one sentence and meaningful in another, which is why removing them needs context, not a blind search.

| Type | Example | Usually cut? |
|---|---|---|
| Hesitation sounds | "um", "uh", "er" | Yes |
| Doubled words | "I, I think" | Yes, keep one |
| False starts | "We built, we shipped it in a week" | Yes, keep the finished version |
| Verbal habits | "so", "basically", "you know" | Often, if it doesn't change tone |
| Hedges | "kind of", "sort of" | Sometimes, if the claim stays true |
| Real words that look like fillers | "it looks like a bug" | No |

## How do I remove filler words automatically?

Use an editor that cuts by transcript word. The AI reads a word-timed transcript, picks the fillers, and deletes those words; the editor converts each word's start and end time into a cut. Because the cut follows word boundaries, it lands between words instead of mid-syllable.

In Vidiyo:

1. Import the clip. Vidiyo transcribes it on your Mac with word timings.
2. Type "cut the ums and uhs" or, for a stricter pass, "remove fillers, doubled words and false starts".
3. The assistant reads the transcript, chooses the words to remove, and deletes them by word ID. It only cuts a filler when the sentence still makes sense without it.
4. Play it back. Each request is one undo step if you want them back.
5. For long recordings, ask "keep only the best take of each line" too. That catches repeated sentences, not only single words.

Fillers and pauses usually go together, so a single request like "cut the ums and long pauses" handles both. Pauses are cut only where the audio confirms silence; see [removing silences on Mac](https://getvidiyo.app/blog/remove-silences-from-video-mac) for how that works.

## Why doesn't AI catch every filler?

AI can only cut a filler that appears in the transcript, and speech recognition doesn't always write every "um" down. [Whisper](https://getvidiyo.app/glossary/whisper-transcription), the open speech model Vidiyo runs on your Mac, often leaves hesitations like "um" and "uh" out of its text, as users discuss on the [official Whisper demo's forum](https://huggingface.co/spaces/openai/whisper/discussions/30). If it isn't in the text, there's no word to delete.

When that happens, the filler is still in the audio. In Vidiyo, move the playhead to it, press B to split on each side, and delete the piece. If a missed filler sits inside a pause, the pause removal may already have taken it, since a very soft "um" can measure as quiet. Uncertain word timings are marked so the assistant cuts on the nearest reliable word rather than guessing.

## How do I remove filler words by hand?

Listen through once and cut each filler you hear, using the waveform to find its edges. Fillers show up as short, low bumps between louder words. It's slow, but you control every cut, and it works in any editor, including free ones.

1. Play the clip at normal speed and drop a marker on every filler you hear.
2. At each marker, zoom in on the waveform until the filler is a distinct bump.
3. Split just before it starts and just after it ends, leaving a couple of frames around neighbouring words.
4. Delete the filler and close the gap.
5. Listen across the cut. If the rhythm feels rushed, extend the gap by a frame or two.

If your editor has a transcript view, search it for "um" and "uh" first. That finds the obvious ones fast; you still need to listen for the ones the transcript dropped.

## Should you remove every filler word?

No. Remove fillers that slow a sentence down or make you sound unsure, and keep ones that carry tone. A video with every hesitation stripped can sound clipped and oddly perfect. One "um" before a joke can be part of the timing. The goal is to sound like yourself on a good day.

A useful rule is the one Vidiyo's assistant follows: cut a filler only when the meaning survives. If removing "kind of" turns a careful claim into an overclaim, leave it. If a false start contains the better phrasing, keep that version, not the finished one.

## How do you say fewer filler words in the first place?

Fillers mostly fill thinking time, so the fix is to think before you speak and let silence do the job. A pause is easy to cut; an "um" in the middle of a word is not. Recording habits reduce fillers more than any editing trick, and they make every cut cleaner.

- **Pause instead of filling.** When you lose your place, stop talking. Silence is free to remove.
- **Use bullet points.** Knowing your next point reduces "so, um, what else".
- **Restart, don't repair.** If a sentence goes wrong, pause and say the whole sentence again.
- **Slow down slightly.** Most fillers come from talking faster than you're thinking.
- **Do a warm-up take.** The first minute usually has the most fillers; plan to cut it.

You won't get to zero, and you shouldn't try on camera; you'll sound stiff. The goal is fewer fillers in the takes you keep. More in [tips for recording yourself on camera](https://getvidiyo.app/blog/record-yourself-on-camera-tips).

## How do you hide the cuts after removing fillers?

Most filler cuts are short enough to pass unnoticed, but some create a visible jump in your position. Change the framing across the cut with a small punch-in, or cover it with a cutaway, and the jump reads as intentional. Avoid adding a transition to every cut; that draws more attention than the jump.

In Vidiyo, ask "punch in on alternate cuts" and it scales one side of the cut slightly with your face kept in frame. Its quality checks also flag raw [jump cuts](https://getvidiyo.app/glossary/jump-cut) with identical framing on both sides. For the full picture, read [how to edit a talking-head video fast](https://getvidiyo.app/blog/edit-talking-head-video-fast).

## Honest limits

- A filler can only be cut by word if the transcript contains it. Anything missed can be cut by hand.
- Hindi transcribes in Devanagari, so Hinglish fillers may need fixing by hand.
- Timelines max out at 5 minutes.

## Frequently asked questions

### How do I remove ums and uhs from a video automatically?

Use an editor with a word-timed transcript. In Vidiyo, type "cut the ums": the assistant finds the fillers in the transcript and deletes those words, and Vidiyo turns that into exact cuts with padding around the neighbouring words.

### Should I remove every filler word?

No. Remove the ones that slow the sentence down, and keep the ones that carry tone or meaning. "Like" in "it looks like a bug" isn't a filler. A few natural ums can also make a video feel less robotic.

### Why didn't the AI catch every um?

It can only cut a filler that appears in the transcript, and speech recognition doesn't always write every um or uh down. Missed ones can be split and deleted by hand at the playhead.

### Does removing filler words create jump cuts?

Yes, each cut inside a shot is a small jump cut. Short filler cuts are often invisible, but if one shows, a punch-in on one side or a cutaway hides it.

## Related

- [Remove silences and filler words automatically](https://getvidiyo.app/features/silence-and-filler-word-removal)
- [Jump cut](https://getvidiyo.app/glossary/jump-cut)
- [Whisper transcription](https://getvidiyo.app/glossary/whisper-transcription)
- [How to remove silences from a video on Mac](https://getvidiyo.app/blog/remove-silences-from-video-mac)
- [How to edit a talking-head video fast](https://getvidiyo.app/blog/edit-talking-head-video-fast)

## Sources

- [About filler words detection (openai/whisper discussion on Hugging Face)](https://huggingface.co/spaces/openai/whisper/discussions/30) (accessed 2026-10-08)

Download Vidiyo for Mac (free during the beta): https://getvidiyo.app/download