Rough podcast editing
Shorten the long gaps left while searching for words, keeping a natural pause of about half a second.
Guide
Remove Silence cleans up overly long gaps in voiceover, online courses and interviews. The panel offers three thresholds — -50 dB “Strict”, -40 dB “Recommended” and -30 dB “Aggressive” — and lets you set a minimum silence length of 0.1–3 seconds. It is built on ffmpeg's silenceremove: leading silence is cut away entirely, while gaps in the middle and at the end that exceed the setting are not deleted in full but shortened to the pause length you chose.
Updated 2026-09-103 min read
Remove Silence cleans up overly long gaps in voiceover, online courses and interviews. The panel offers three thresholds — -50 dB “Strict”, -40 dB “Recommended” and -30 dB “Aggressive” — and lets you set a minimum silence length of 0.1–3 seconds. It is built on ffmpeg's silenceremove: leading silence is cut away entirely, while gaps in the middle and at the end that exceed the setting are not deleted in full but shortened to the pause length you chose.
That makes it less “shave away every silent sample” and more “squeeze the gaps shorter”. The defaults are -40 dB and 0.5 seconds, and the code warns explicitly that -30 dB can mistake breathing and mouth noise for silence.
| Input | Output | Notes |
|---|---|---|
a 10-minute voiceover with 2–4 second gaps between sections |
-40 dB, 0.5 s → long gaps become pauses of about 0.5 s |
leading silence is removed outright |
a very quiet ASMR recording |
the -50 dB strict threshold |
lowers the risk of treating quiet sound as silence |
Shorten the long gaps left while searching for words, keeping a natural pause of about half a second.
Cut long pauses across a lecture so the result feels tighter.
Denoise first, then remove the leading and between-section gaps with the recommended threshold.
If the “silence” contains an air conditioner, keyboard noise or a noise floor above the threshold, silenceremove will not treat it as silence.
Anything longer than that triggers processing, and after processing about 0.5 seconds of pause is kept in the middle and at the end.
The audio is processed locally in the browser by ffmpeg.wasm running silenceremove and is never uploaded. The first time you open the workbench, a single-threaded ffmpeg core of about 31–32 MB is downloaded from the site asset domain or jsDelivr.
Updated 2026-09-10
silenceremove trims silent stretches at the start, middle and end; threshold and minimum length adjustable
Everything runs locally in your browser; files are never uploaded to a server