Loading page
About this tool
Removes long silences from an audio recording using an adjustable loudness threshold.
The Silence & Filler-Word Cutter runs FFmpeg's silenceremove filter over an audio file in your browser, deleting any stretch that stays below your chosen loudness threshold for longer than 0.65 seconds. You pick the threshold — -35 dB, -40 dB or -45 dB — upload the recording, and get back a WAV with the dead air taken out and the speech left untouched. It is aimed at podcasters, course recorders and anyone whose raw take is padded with pauses they would otherwise trim by hand.
Open Silence & Filler-Word Cutter on AltFTool — it loads instantly in your browser.
Choose your recording with the Source file picker — any audio file type is accepted.
Pick a Silence threshold of -35dB, -40dB or -45dB and press Process locally; the FFmpeg WebAssembly engine loads only then and deletes stretches quieter than the threshold for longer than 0.65 seconds.
The trimmed audio downloads automatically as altftool-silence-filler-word-cutter.wav, and the Local processing report panel shows the file details and FFmpeg log.
Three explicit levels (-35, -40, -45 dB) let you match the cut to your actual noise floor instead of trusting a hidden auto-detect.
Only silence lasting more than 0.65 seconds is removed, so natural breaths and beat-pauses between sentences stay in and speech does not sound clipped together.
Processing runs on an FFmpeg WebAssembly build inside the tab, so unreleased interviews and client recordings are not uploaded anywhere.
No — it removes silence, not fillers. Cutting an "um" reliably needs transcript timestamps that a human has reviewed, and guessing at them from the waveform alone would take out real words, so filler removal is left to a reviewed edit.
Longer than 0.65 seconds below your chosen threshold. Anything shorter is kept, which is why normal breaths and sentence gaps come through intact.
Start at -40 dB for a reasonably quiet room. Use -35 dB if pauses are surviving because of a high noise floor (fan, street, mic hiss), and -45 dB if speech tails or quiet words are being clipped off.
A WAV file, regardless of what you put in. WAV is uncompressed, so the output is larger than an MP3 input but loses nothing before you take it into an editor for the rest of the work.