Updated October 2026 · Editing audio well
Remove filler words such as um and er where they distract, but keep the ones that give the speech its natural feel. Cut in small amounts, avoid joins that change the rhythm, and make sure the speaker still sounds like themselves. Over-removal makes people sound robotic and rehearsed.
Filler words are the small sounds people make while thinking: um, er, uh, and phrases such as you know. In a transcript they can look like clutter, and many editors remove them to tighten the speech. Used carefully, that removal makes an episode easier to follow. Used heavily, it turns a real person into a polished voice that no longer sounds like the speaker, and listeners often notice the difference even if they cannot say why.
This post explains which fillers to cut, which to keep, how to make the joins invisible and how to keep the speaker recognisable.
Fillers signal thought. A short um before an answer can show that the speaker is considering the question, and it makes the answer feel more natural. Removing every filler can make an answer sound as though it was prepared in advance, which may not be what the listener expects from a conversation.
A filler becomes a problem when it is repeated, when it breaks the flow of an important point or when it distracts from the content.
Listen to the speaker and note the fillers. Group them into three kinds: the ones that distract, such as a long string of ums in a row; the ones that carry thought, such as a short pause with a filler before a key idea; and the ones that are part of the speaker’s style. Cut the first group, keep the second, and decide the third by listening.
Remove fillers one at a time, not in batches. After each cut, listen to the sentence at normal speed. A filler removed with a short fade and a matched breath often disappears completely, while a batch of removals can make the speech sound jerky. If a join is noticeable, restore a little of the filler to soften it.
When a filler is cut, the breath around it often needs adjusting. Keep a natural breath at the join, and make sure the rhythm of the sentence still sounds like the speaker. A sentence with its fillers removed can end up too fast, and a little pause restored at the join may fix that.
Each speaker has a rhythm and a style. A speaker who uses filler words as part of their conversational voice may sound unlike themselves if every filler is removed. Listeners who know the speaker will notice, and they may feel the episode is less honest. Keep enough of the style that the speaker is still recognisable.
Automatic transcripts often show fillers clearly, which makes them easy to find. Use the transcript to locate the fillers, but always make the cut in the audio, and listen to the result. A transcript can show a filler as removed while the audio still carries a click or a change in room tone.
Some speakers want the fillers gone, and others feel that their natural voice is part of the show. Ask before you start, and agree on a level of editing that suits the speaker and the audience. A short note of the agreed approach at the top of the project keeps later edits consistent with what the speaker expects. Playback at the speaker's own pace is the best final check.
How much to cut depends on the format and the audience. A tightly produced show may cut more than a relaxed conversation, and a formal interview may cut more than a casual chat. There is no fixed number of fillers to remove. The aim is clear speech that still sounds like the speaker.
Editing a speaker’s words, even lightly, should be agreed in advance. Some guests expect a tidy edit, and others want their conversational style kept. Discuss the approach before recording, and confirm it before publishing, as set out in the consent post in this series.
A final listen on a phone speaker shows whether the speech still flows. The Vocal Checker shows clipping, noise and level problems in a recording you own, and the Tunedzy Go page covers cleaning phone recordings. For the transcript side of the work, read the show notes post in this series.
Spot clipping, noise, rumble and level problems. Runs in your browser, nothing is uploaded.
No. Remove the distracting ones and keep the fillers that carry thought or the speaker’s style.
Fillers were cut in batches. Cut one at a time, with a breath and a short fade at each join.
Use it to locate them, then make every cut in the audio and listen to the result.
Agree the editing approach with guests before recording or publishing, especially for interviews.
