Why the free tool cannot do this
The tool on this site measures loudness. An “um” is as loud as any other word, so only a speech recognizer that hears the word can find it.
That is the difference between the two jobs. Silence removal looks for gaps in the level. Filler word removal transcribes the recording, labels each word with its start and end time, and cuts the ones that are hesitations. Most tools that do the second also do the first, because the pauses around an “um” are usually worth cutting too.
Remove filler words with QuellCut
- Import the video or audio into QuellCut.
- Open and choose the language spoken in the recording.
- Turn on . Turn on to keep only the final take of a repeated sentence.
- Review the cuts on the timeline, then export.
QuellCut uses Apple speech recognition, on the device wherever your language supports it.
Repetitions are a separate problem from fillers. When a speaker starts a sentence, stops and says it again, both versions are clear words, so neither looks like a filler. QuellCut flags repeated phrases and false starts so you can keep the last, cleanest take. Every cut stays on the timeline, where you can turn it off or adjust its edges before you export the full video, vertical Shorts or audio as M4A or WAV.
In Premiere Pro
Premiere Pro’s transcript can mark filler words alongside pauses. Adobe’s current help pages call the app Adobe Premiere; the feature is part of Text-Based Editing and is included with the subscription.
- Transcribe the clips. In the panel, select , or turn on automatic transcription when you import.
- Select the panel so the Text panel switches to the sequence transcript. Bulk deletes work there.
- Select the filter icon in the panel and choose . Every match is highlighted.
- Select for one match or for every match. closes the gap, leaves it.
Review each one before you delete it: a highlighted word can be a real word, and a long run of deletions can leave choppy video. Clips transcribed before filler detection was added need to be transcribed again before fillers show up. For the pause side of the same workflow, see remove silence in Premiere Pro.
In CapCut desktop
CapCut flags fillers through its captions. According to CapCut’s own page, you open , choose , pick the language, turn on the filler word option and select . The flagged words are highlighted in the transcript, where you delete them. The option’s label and whether it needs CapCut Pro have changed between versions, so check yours. Silence removal in CapCut is covered in remove silence in CapCut.
Tools compared
| Tool | Finds fillers | Also removes pauses | Repeated takes | Platform |
|---|---|---|---|---|
| Browser tool on this site | No | Yes | No | Any browser |
| QuellCut | Yes, automatic | Yes | Yes, flagged for review | iPhone and iPad |
| Premiere Pro | Yes, from the transcript | Yes | No | Mac and Windows |
| CapCut desktop | Yes, from captions | Varies by version | No | Mac and Windows |
| Audacity | No | Yes, Truncate Silence | No | Mac, Windows, Linux |
Common mistakes
- Deleting real words. “Like”, “so” and “right” are fillers in one sentence and meaning in the next. Review before a bulk delete.
- Wrong language. A recognizer set to the wrong language hears noise instead of words, and the fillers are missed. Choose the spoken language first.
- Cutting too close. A filler is often glued to the next word. A cut with no padding can take the first sound of that word. Listen to each join.
- Visible jumps in video. Each removed “um” is a jump cut. On a locked-off shot that is normal for YouTube, but in an interview it can look busy. Cover the worst joins with B-roll or leave a few fillers in.
- Noisy audio. Recognition gets worse with background noise, music and crosstalk. Clean audio gives better detection in every tool.
When to keep filler words
Fillers are not always mistakes. In an interview, a short “um” before an answer tells the listener the guest is thinking, and cutting it can make a reply sound rehearsed. In a podcast with two hosts, fillers help mark who is about to speak. Removing every one can make a conversation sound edited in a way listeners notice.
A practical rule: remove fillers that interrupt a sentence, that come in a cluster, or that sit at the start of a clip you will post on its own. Keep the occasional one that sounds natural. Every tool above lets you review and skip matches, so use that instead of a blind bulk delete.
How to check the result
Play the edit at normal speed from start to finish. Listen for sentences that now run together without a breath, and for words that begin too abruptly. If you are counting fillers to work on your delivery, filler words per minute explains how to measure them.
If you only need the pauses gone
When the speaker is fluent and the problem is dead air, you do not need speech recognition. The free browser tool cuts silence from audio in your browser, and remove silence from video covers the editors that cut picture too. For a background on why long gaps hurt a recording, see what dead air is.
Questions
- Can a free silence remover take out um and uh?
- No. Silence removers cut audio that is quieter than a threshold, and a filler word is as loud as normal speech. You need a tool that transcribes the recording.
- Which filler words are detected?
- Most tools focus on hesitation sounds such as um and uh. Adobe describes Premiere’s detection as covering uh and umm. Words like “like” or “so” are usually real words too, so treat them by hand.
- Should I remove every filler word?
- Usually not. A few hesitations sound human, and removing all of them can make speech feel clipped. Aim to remove the ones that break a sentence or come in clusters.
- Does the recording have to be in English?
- No, but support depends on the tool. QuellCut uses Apple speech recognition and asks for the spoken language. Premiere’s Speech to Text supports a list of languages, and Adobe says filler detection works across them.
- Is my recording sent to a server?
- In QuellCut, no: recognition runs on the device wherever the language supports it. For desktop editors, check the vendor’s own documentation for your version.
Sources: Adobe Help: detect and delete pauses, Adobe Help: Text-Based Editing overview, Adobe Community announcement, CapCut: filler words.
