How does this audio silence remover decide what to cut?

A quiet region is a continuous run of samples at or below your dBFS threshold in every channel. It must reach the minimum quiet duration, and enough samples must remain after padding to make a cut. The defaults are -40 dBFS, 500 ms minimum duration and 100 ms retained on each side. This detects low amplitude, not semantic silence: quiet speech, music, breaths and background sound can qualify. Listen to each suggestion and uncheck Remove to keep it.

How much audio does padding preserve?

Padding is retained at both ends of each qualifying quiet region, including regions at the beginning or end of the file. A one-second quiet region with 100 ms on each side loses exactly 0.8 seconds. With 250 ms per side it loses 0.5 seconds. If the two padding intervals cover the entire region, no cut is proposed. Padding rounds to the nearest decoded sample; the minimum duration rounds up to a whole sample.

Can I preview and reverse individual cuts?

Yes. The table gives original quiet-region times, proposed cut times and the duration selected for removal. Audition starts up to a quarter-second before the region and stops shortly after it; the original player also supports seeking. Uncheck Remove to keep a region, or use Keep all regions. The retained-region list and duration totals update with your decisions. Prepare edited audio to listen to the selected edit and enable Download WAV. A settings change clears the old analysis and decisions; a decision change clears the prepared edit so you cannot download stale cuts.

What happens if the whole recording is quiet?

The tool flags a recording whose samples are all at or below the threshold. Review the original because low-level speech or music may still be present. Padding can retain audio at the edges. If the selected cuts would remove every sample, preview and download are disabled until you keep a region or increase padding. The tool does not export an empty WAV.

Which audio files and sizes can I use?

Choose a local WAV, MP3, M4A, OGG, FLAC or AAC file supported by your browser. The file limit is 50,000,000 bytes. A file must contain one mono or stereo audio track at 8,000 to 192,000 Hz, with no format changes or discontinuous timestamps. Decoded audio is limited to 10 minutes and 32 million channel samples, whichever is reached first. Analysis allows up to 2,000 proposed cuts and each processing job stops after two minutes. Larger recordings need to be split or converted before retrying. Unsupported or damaged files produce an error and leave your chosen file and settings available.

Is the download an MP3 or a WAV?

The download is uncompressed 16-bit PCM WAV at the decoded sample rate, with the original mono or stereo channel count. It contains the retained regions in their original order. Preview and download use the same prepared WAV. Samples outside full scale are clipped and float samples are quantized to 16 bits. There is no crossfade between joins, so a cut can click. A WAV may be larger than a compressed source.

Is my audio uploaded, and can I cancel processing?

Audio decoding, analysis and WAV preparation run locally in a worker in this tab. Your audio is not uploaded. Cancel processing terminates the worker and keeps the chosen file and settings for a retry. Choosing a different file invalidates previous results. Leaving the page releases the worker and local preview URLs; files and settings are not saved between visits.

Audio Silence Remover

Find quiet regions in a local audio file up to 50 MB, then decide which pauses to remove. Set the threshold, minimum quiet duration and padding, audition the original, keep individual regions, and preview the selected edit before downloading a WAV.

Local audio · review every cutUp to 50 MB

Choose audio

Threshold (dBFS)Minimum quiet region (ms)Keep on each side (ms)

A region qualifies only while every channel stays at or below the threshold. Quiet speech or music may be removed. Changing settings clears cut decisions.

Find the pauses worth cutting.Your waveform and suggested cuts appear after analysis.

Choose a local audio file, then analyze it to review suggested cuts.

OriginalAwaiting audio

Selected removalNot analyzed

After editsNot analyzed

Cut regionsNot analyzed

Local files up to 50 MB (50,000,000 bytes). Mono or stereo, 8–192 kHz, at most 10 minutes and 32 million decoded channel samples. Processing stops after two minutes. Supported codecs depend on your browser. Nothing is uploaded.

Worked examples

Keep breathing room1.00 s quiet; 100 ms per side

0.80 s removed; 0.20 s retained

Increase padding1.00 s quiet; 250 ms per side

0.50 s removed; 0.50 s retained

Above threshold2.00 s at -20 dBFS; threshold -40 dBFS

0.00 s removed; 2.00 s retained

Accuracy. Quiet speech or music can be classified as silence: detection uses sample amplitude across every channel, not meaning. Every proposed cut is inspectable and reversible, with padding retained on both ends of each quiet region. Export is 16-bit PCM WAV at the decoded sample rate; an empty edit cannot be exported. Joins have no crossfade and can click; audition the edit before exporting.

Common questions

How does this audio silence remover decide what to cut?: A quiet region is a continuous run of samples at or below your dBFS threshold in every channel. It must reach the minimum quiet duration, and enough samples must remain after padding to make a cut. The defaults are -40 dBFS, 500 ms minimum duration and 100 ms retained on each side. This detects low amplitude, not semantic silence: quiet speech, music, breaths and background sound can qualify. Listen to each suggestion and uncheck Remove to keep it.
How much audio does padding preserve?: Padding is retained at both ends of each qualifying quiet region, including regions at the beginning or end of the file. A one-second quiet region with 100 ms on each side loses exactly 0.8 seconds. With 250 ms per side it loses 0.5 seconds. If the two padding intervals cover the entire region, no cut is proposed. Padding rounds to the nearest decoded sample; the minimum duration rounds up to a whole sample.
Can I preview and reverse individual cuts?: Yes. The table gives original quiet-region times, proposed cut times and the duration selected for removal. Audition starts up to a quarter-second before the region and stops shortly after it; the original player also supports seeking. Uncheck Remove to keep a region, or use Keep all regions. The retained-region list and duration totals update with your decisions. Prepare edited audio to listen to the selected edit and enable Download WAV. A settings change clears the old analysis and decisions; a decision change clears the prepared edit so you cannot download stale cuts.
What happens if the whole recording is quiet?: The tool flags a recording whose samples are all at or below the threshold. Review the original because low-level speech or music may still be present. Padding can retain audio at the edges. If the selected cuts would remove every sample, preview and download are disabled until you keep a region or increase padding. The tool does not export an empty WAV.
Which audio files and sizes can I use?: Choose a local WAV, MP3, M4A, OGG, FLAC or AAC file supported by your browser. The file limit is 50,000,000 bytes. A file must contain one mono or stereo audio track at 8,000 to 192,000 Hz, with no format changes or discontinuous timestamps. Decoded audio is limited to 10 minutes and 32 million channel samples, whichever is reached first. Analysis allows up to 2,000 proposed cuts and each processing job stops after two minutes. Larger recordings need to be split or converted before retrying. Unsupported or damaged files produce an error and leave your chosen file and settings available.
Is the download an MP3 or a WAV?: The download is uncompressed 16-bit PCM WAV at the decoded sample rate, with the original mono or stereo channel count. It contains the retained regions in their original order. Preview and download use the same prepared WAV. Samples outside full scale are clipped and float samples are quantized to 16 bits. There is no crossfade between joins, so a cut can click. A WAV may be larger than a compressed source.
Is my audio uploaded, and can I cancel processing?: Audio decoding, analysis and WAV preparation run locally in a worker in this tab. Your audio is not uploaded. Cancel processing terminates the worker and keeps the chosen file and settings for a retry. Choosing a different file invalidates previous results. Leaving the page releases the worker and local preview URLs; files and settings are not saved between visits.

Quiet speech or music can be classified as silence: detection uses sample amplitude across every channel, not meaning. Every proposed cut is inspectable and reversible, with padding retained on both ends of each quiet region. Export is 16-bit PCM WAV at the decoded sample rate; an empty edit cannot be exported.