How to Cut Long Pauses and "Um/Uh" Filler Words from Recordings

Spoken recordings naturally include a few seconds of thinking pauses and filler sounds like "um" and "uh". You don't notice them in conversation, but on playback they drag. That's why paid editing services advertise automatic filler and silence removal. The FreeSoundTools Silence & Filler Word Cutter does the same job in your browser, without uploading the file.

What the tool looks for

• Long silences: it measures loudness in very short slices (10 ms) and finds places where it stays quieter than the background noise for longer than a set time (0.8 s by default). Separate on and off thresholds keep the boundary steady where speech stops and starts.
• Filler candidates: short (about 0.1 to 0.7 s), steady sounds like "ummm" with a pause on both sides are marked as candidates. Short answers like "yeah" or "ah" can look similar, so candidates start unchecked.

Speech recognition models often drop filler words from transcripts, so this tool doesn't use a model; it looks only at the shape of the sound. That makes it fast and language-independent, but you should listen to the candidates before cutting.

How to use it

1. Add a recording, or press "Send to filler cut" in the FreeSoundTools recorder.
2. The silences and filler candidates it found appear in a list and as colored marks on the waveform. Press ▶ to hear each one with 1 second before and after.
3. Check only what you want to cut. Silences are checked by default; listen to filler candidates and pick them yourself.
4. Use "Preview result" to hear the whole thing, then save as WAV or MP3. Every cut gets a very short crossfade so there's no click.

Tips for natural-sounding cuts

• Don't remove every pause. A 0.2 to 0.3 second pause between sentences gives listeners time to process. Setting "Padding to keep" to about 0.25 s sounds natural.
• Use a longer threshold for lectures. If there's a lot of writing on the board or slide changes, set "Silences longer than" to 1.5 s or more so needed pauses aren't cut.
• Check pauses after questions in interviews. A thoughtful pause can be part of the answer's meaning. Uncheck those to keep them.
• Avoid recordings with background music. Continuous music makes silences hard to find. Work on the voice-only original and add music afterwards.

Using it with other tools

For noisy recordings, clean them up with Studio Sound first so silences are detected more accurately. After cutting, run Audio to Text for a cleaner transcript with fewer fillers. If you only need to tidy the start and end, Fade In/Out & Remove Silence is simpler.

How much shorter does it get?

It depends on speaking habits, but solo lectures and explainer recordings often get 10 to 20% shorter just by removing long silences. In a 17-second test recording, the tool found 3 long silences and 2 filler candidates and brought it down to about 12 seconds. Cutting too much leaves no room to breathe and becomes harder to follow, so listen to the result from start to finish before saving. If you don't like it, the original is still there; change the checks and make it again. If you clean up the same speaker's recordings regularly, note the "Silences longer than" and "Padding to keep" values that worked and reuse them for consistent results.

In podcasts with two or more people, the short gap as one person finishes and the other responds is the rhythm of the conversation. For those, set the silence threshold to 1 second or more and only clean up the truly long breaks.

Open the Silence & Filler Word Cutter

Frequently asked questions