Remove silence from audio

Cut out stretches of an audio file (or a video's audio track) quieter than a threshold for longer than a minimum duration, keeping a little padding at each cut, decoded and re-encoded as WAV through the same PCM path as audio-reverse. WAV decodes everywhere; other codecs need a browser.

Remove silence from audio cuts any stretch of low volume once it runs longer than a minimum length you set, leaving a short pad near the nearest real sound. Loudness is measured against a threshold in dBFS, minus 40 by default. A file quiet throughout is refused. WAV decodes anywhere; other codecs need a browser first.

Questions

How is silence detected?

Every sample is compared against thresholdDb, minus 40 dBFS by default, converted to a linear level; a frame quieter than that on every channel counts as silence.

What does paddingMs do?

It keeps a short stretch of the removed silence, 0 milliseconds by default, at whichever edge of a cut run borders real audio, so the cut does not land jarringly close to speech.

What if there is no silence to remove?

The tool still writes an output file, a copy of the input re-encoded as WAV, and its report says no run of silence at or above minSilenceMs was found.

Why did it refuse my file entirely?

Because the whole file measured below thresholdDb, so removing all of it would leave nothing. Raise thresholdDb or check that you dropped the right file.

Does this work on a video's audio track?

Yes. The picture is untouched; only the decoded audio samples are trimmed and written back out as a WAV file.

Related Video tools

n→nConvert audioChange an audio file between MP3, WAV, FLAC, Ogg and M4A (AAC). In the browser only, like video-convert, because mediabunny decodes and encodes through WebCodecs, which Node has no equivalent of.n→nNormalize audio volumeScale an audio (or video's audio track) so its loudest sample hits a target peak level: a two-pass peak scan followed by a gain-scaled re-encode, both through WebCodecs, so this runs in the browser only, like video-convert. Always outputs MP4 with an unchanged video track (if any).n→1Trim or merge audioCut an audio file down to a start-end range, or concatenate two or more audio files one after another. A packet-level copy with no decoding, so this runs in Node, the CLI and MCP as well as the browser. Output is always an MP4 audio file (.m4a).n→nExtract audio from videoPull the audio track out of a video into its own MP4 audio file. A packet-level copy with no decoding or re-encoding, so this runs in Node, the CLI and MCP as well as the browser.n→nMute videoDrop the audio track from a video, keeping the picture untouched. A packet-level copy of the video track only, no decoding, so this runs in Node, the CLI and MCP as well as the browser.n→nReverse audioPlay an audio file (or a video's audio track) end to start, decoded to raw samples and re-encoded as WAV. WAV input decodes in Node, the CLI and MCP as well as the browser, the same PCM path audio-trim-merge's packet copy sits next to; AAC, MP3, Opus, FLAC and Vorbis need a browser.n→nFade audio in or outRamp an audio file's (or a video's audio track's) volume up from silence at the start, down to silence at the end, or both, linear or logarithmic, decoded and re-encoded as WAV through the same PCM path as audio-reverse. WAV decodes everywhere; AAC, MP3, Opus, FLAC and Vorbis need a browser.n→nConvert mono to stereo or backChange an audio file's (or a video's audio track's) channel count: average every channel down to mono, duplicate a mono track into identical left and right channels, or swap left and right, decoded and re-encoded as WAV through the same PCM path as audio-reverse. WAV decodes everywhere; other codecs need a browser.