How to Clean Filler Words Before a Voiceover (2026)

Cleaning filler words before a voiceover means you edit the lines someone is about to read, or the text you are about to treat as that script. You are not polishing a transcript for its own sake.

Um, uh, "you know," and a false start make a live person sound alive. They make a booked read stumble, burn booth time, and leave the talent unsure which half of the sentence was the real one. The fix is a pass with judgment. A blind delete of every filler will also delete the hesitation that is the story, and it will mangle ordinary words that only look like filler.

This pass sits next to how to turn a transcript into a voiceover script. Shape the script first if you have not chosen the lines yet. Clean fillers on what remains. The wider case, publishing, profanity, and a transcript that is not about to be performed, is enhancing transcriptions by removing filler words and profanity.

Get the words from the recording by uploading it in the browser on the homepage. File Transcribe is not an offline app and not a Windows install. Guest try covers a first file with no signup. Free is about 50 minutes a month. Pro is about $19 a month for about 2,000 minutes. Confirm on pricing.

Direct answer

1. Work from the script lines you are actually recording, with timestamps still attached.
2. Delete fillers that make the reader trip: stray um, uh, "you know" with no referent, and abandoned starts you do not want played.
3. Keep a filler when the hesitation is the point, as in "I, um, quit."
4. Do not run a find-and-replace that deletes every um, like, or you know.
5. Read the line aloud once. If you stumble, the line is not clean yet.

Two different ums

One um is throat-clearing. The speaker is finding the next clause. In a product read, a course voiceover, or an ad, that um costs a breath and adds nothing the picture needs. Take it out.

The other um is the content. The person is about to say something they can barely say. Strip it and you have a different story, delivered with a confidence they did not have.

"I, um, quit."

Leave it if the voiceover is that scene. Tell the talent the break is intentional, or write it as a beat: "I... quit." or "I [beat] quit." A cleaned "I quit" is a better commercial and a worse documentary. Pick the piece you are making before you touch the word.

"You know" and "like" are sometimes real words

"You know" in the middle of a clause is often empty. "You know, we ship on Friday" can become "We ship on Friday." The listener was never being asked.

"You know the password" is a different sentence. Delete "you know" there and you have "the password," which is not what they said and not what anyone should read on air.

"Like" has the same trap. "It tastes like metal" must stay. "I was, like, so late" is filler. For a clean commercial voice, cut it. For a character read that is supposed to sound like that person, keep it.

Um and uh are safer targets than like, well, so, and right. "So you open the log" is a real transition. Cutting every "so" leaves a stack of commands with no joints.

False starts: keep the fact, drop the abandoned try

A false start is the speaker leaving a sentence on the floor.

"The price is fifty. The price is fifteen dollars."

If the script is a product read, the line is "The price is fifteen dollars." The fifty was a mistake they corrected. Reading both will make the talent, and the viewer, remember the wrong number. Put the timestamp beside the fifteen so you can prove the correction was theirs.

If the piece is about them catching the mistake, the false start stays, and you mark it so nobody "helps" by deleting it in the booth: "The price is fifty. [they correct] The price is fifteen dollars."

Do this after the script pass has already thrown away the meeting preamble. Cleaning a false start inside "can you hear me" is wasted work. That cut belongs in how to turn a transcript into a voiceover script.

Do not run a blind delete-all

A replace of "um" can hit a note you typed or a careful "I, um, quit" you meant to protect. A replace of "like" guts comparisons. A replace of "you know" guts questions. You will notice in the booth, which is late.

Go line by line on the pages you are recording. If the document is still twenty pages, you are cleaning a transcript. Go back to how to turn a transcript into a voiceover script and cut first.

A one-click cleaner in a video editor has the same blind spot. Read every changed line before anyone books time. Descript is the name people reach for when the transcript is also the edit. Confirm what it deletes on descript.com.

Hand the talent a line they can say once

After the cuts, the line should survive a single cold read. If the margin says "skip that word," delete the word on the page. Margins are for pronunciation and for intentional beats.

Leave the source timestamp on the block. "About fifteen dollars" and "fifteen dollars" are not the same promise. This pass does not get to upgrade a hedge into a guarantee.

FAQ

How do I clean filler words before a voiceover?

Cut um, uh, empty "you know," and false starts that make the reader stumble. Keep a filler when the pause is the meaning. Read the line aloud before you record.

Should I delete every um?

No. Delete the ones that are only delay. Keep "I, um, quit" if the hesitation is the scene. A global delete cannot tell those apart.

What about "like" and "you know"?

Delete them when they carry no meaning. Keep them when they are the grammar of the sentence, or when you are casting that person's voice on purpose.

Is this the same as cleaning a transcript for a blog?

No. A public page, a profanity pass, and a general readability edit are the older guide, enhancing transcriptions by removing filler words and profanity. This page is only the read.

Upload the recording on the homepage, shape it with how to turn a transcript into a voiceover script, and clean the remaining fillers on the lines you are about to record.

More guides

• Transcription guides
• Try File Transcribe free
• Transcript format guide
• Best transcription software
• Transcribe Zoom meetings

Further reading

• Zoom cloud recording help
• Google Speech-to-Text docs