Guides › Editing

How to remove ums, uhs and long pauses from a recording

“Um”, “uh”, a sentence you started twice, a long silence while you found the next thought: in conversation nobody notices, but in a podcast or a voiceover they make you sound slow and unsure. Here is what to cut and what to keep, how to edit them out by hand without audible joins, and how to find them all at once from the transcript and listen before anything is cut.

Don't want to hunt for every um? Remove them from the transcript with Respeak, free →

What to cut, and what to keep

• Fillers: um, uh, er, ah, hmm. When they stand on their own while you think, they can almost always go.
• Repeated words: “I I think”, “the the”. Keep the last one.
• Restarts: “we're going to, we're going to talk about…”: cut the start you abandoned.
• Long pauses: silences over a second can usually be shortened, but not removed altogether.

Keep: normal breaths, pauses you left on purpose for emphasis, and words like “like” or “so” when they carry meaning. Taking every last one out sounds robotic; cut until it no longer drags.

By hand: cutting in an audio editor

1. Find it: listen with the waveform in view; a filler is usually a small separate blob between two phrases.
2. Cut in the silence: after the previous word ends and before the next one starts. Cutting into a word clips its start or end.
3. Add short fades: 10–20 ms at each cut hides the click at the join.
4. Leave some pause: after removing a filler, keep 0.3–0.5 seconds of the silence around it so the phrases don't collide.
5. Listen through: cutting too much sounds rushed; the obvious ones are enough.

The trouble is time: ten minutes of talk can hold dozens of them, each to be heard, found and cut.

From the transcript: Remove fillers in Respeak

Remove fillers turns the recording into a transcript; you delete words there and the audio follows:

1. Upload the recording, transcribe it and click Remove fillers at the top of the transcript. The first time, it times every word of the recording: under a minute for a short one, a few minutes for a long one.
2. Fillers are highlighted, possible repeats and restarts are underlined, and pauses over a second show as “pause” chips. Fillers the transcript left out but the recording has are put back in grey, so you can cut them too.
3. Delete all fillers or shorten all long pauses in one click (pauses kept at 0.3–0.8 seconds, your choice), or Review one by one: each one plays, Enter deletes, S keeps, ← → move.
4. Click any other word to delete it, or drag across several; ▶ plays a sentence as it will sound, and Ctrl+Z takes back the last step.
5. Apply cuts and the recording and transcript update together. Undo if you don't like it; export WAV or MP3 when you do.

Remove fillers is free; the first timing counts toward the day's transcription time. And if a sentence turns out to be wrong, you can change it in your own voice right there: 3 free generations before each replacement, 1 credit only when you replace.

Try it on your own recording →

Limits: Chinese (Mandarin), English, Japanese, Korean, Spanish, French, German and Portuguese. Loud music under the voice or several people talking at once make the transcript less accurate. What is marked is only a suggestion: some may be missed or wrong, so listen before you cut.

Common questions

Will it remove “like” and “you know” too?

Not by itself: those are often real words, so they are not suggested. Click them in the transcript to delete the ones you don't want.

How long should a pause be?

Between sentences, 0.3–0.6 seconds usually sounds natural: shorter for fast-paced talk, longer for storytelling or where you want a beat. Respeak keeps 0.45 seconds by default, and you can choose 0.3, 0.6 or 0.8 seconds.

It sounds rushed after cutting. What now?

Keep pauses a little longer (0.6 seconds, say), or only cut the most obvious fillers and leave some in. If you don't like the result, undo it in one click and cut again with other settings.