No. Reading, analysis and export all happen inside the browser. The result leaves only as a download you start yourself; neither the original nor the processed file is sent to a server. Unreleased recordings, internal meetings and interview material you are not free to share are all fine to run through it.

How long a file can it handle?

Up to about an hour is usually comfortable. The size ceiling is 200 MB, and anything larger is refused as you load it. Long files take a while to analyse, but progress is reported as a percentage rather than leaving the page apparently frozen. Note that the audio is expanded into memory to be examined, so on a modest device even a file under the limit can feel heavy — split it first if so.

The output file is bigger than the original

That happens when you save as WAV. WAV is uncompressed, so exporting a compressed source such as an MP3 grows the file even though the duration shrank; stereo WAV runs to roughly 10 MB a minute. Choose MP3 to keep the size comparable. If you plan to keep editing afterwards, though, staying with WAV avoids a second round of lossy compression.

Can it also raise the level or reduce noise?

No — this one only shortens the timeline. Level and tone belong to the mastering tool, separating vocals from backing to the stem separator, loudness measurement to the key & BPM analyser, key and tempo to the key & tempo changer, and format conversion to the file converter. The intended flow is to close the gaps here, save as WAV, and take that to whichever page does the next job. To make the recording in the first place, there is a voice recorder page.

Can I see where the cuts were made?

Yes. Below the figures, the whole file is drawn as a single strip: left edge is the start, right edge the end, and the shaded areas are what came out. Seeing whether the silence was all at the front or scattered throughout is what tells you which setting to adjust. The preview player underneath lets you hear the result directly.

The threshold is your definition of quiet

0 dB is the top of the scale and lower numbers mean quieter: −60 dB is close to true silence, −40 dB is about the noise floor of a quiet room. If only digital silence should count, start near −60 dB; for a recording with air conditioning, fan noise or tape hiss, −50 to −40 dB is the useful range. When detection misses, move this figure in 5 dB steps before touching anything else. Push it too high and quiet speech and breaths start being judged silent — a sudden jump in the number of gaps found is the warning sign. The measurement itself is an average over roughly twelve milliseconds rather than an instantaneous peak, so a single knock or a burst of equipment noise does not break a silence in two — at the cost of reacting less to very short sounds.

The minimum silence stops it over-tightening

Quiet stretches shorter than this are kept even when they fall below the threshold. Lower values close smaller gaps and shorten the file further, but below about 0.2 seconds the spaces between words and around punctuation start to go, and the same content comes back sounding hurried. For conversation and narration, 0.4 to 0.8 seconds is the comfortable range. If you only want to lose the waiting around at the start and between takes, set it to two seconds or more and the delivery is left completely alone.

Do not set the padding to zero

Cutting a silence exactly to its edges eats the attack of the next word and the tail of the last. Both are quiet, and both matter far more to comprehension than their level suggests, so the result reads as breathless rather than shorter. Keeping the default hundred milliseconds at each end is what makes the edit go unnoticed. If you need to remove more, shorten the minimum silence rather than the padding — it costs less in listenability.

It is not suited to instrumental music

The test is level alone, so it cannot tell a musically meaningful silence from empty space. A decaying note, the middle of a fade, a held pause between phrases — any of them can fall below the threshold and be treated as a gap. This is a tool for material where sound is either present or absent: talk, narration, lectures, interviews. On music, set the minimum silence to two seconds or more so only the spaces between pieces qualify, and always check the preview.

Back to all tools／ Music & audio

Silence remover

Runs in your browserNothing is uploaded

Finds the stretches where nothing is happening and closes them up. Set the threshold, the shortest silence worth cutting and how much padding to keep, then see the before and after lengths, how much came out, and how many gaps were found.

Input

What this tool decides

Threshold: -45 dB
Shortest silence: 0.50 s
Padding kept: 100 ms

Drop in some audio and these three become sliders. Move any of them and the map of what would be cut, and the resulting length, update straight away — before you commit to anything.

Output

The tightened audio lands here, with the before/after length and how much was cut.

Cutting on a threshold alone swallows breaths and the gaps between words, which sounds wrong. That is why a minimum silence length and a margin at each join can both be set.

All tools

Highlights

Hands the judgement-free edit to the machine: An hour of recorded material with forty minutes of actual talking is ordinary. Selecting and deleting the rest by hand contains nothing that deserves to be called a decision. Settle two questions — how quiet counts as silence, and how long it has to last — and the rest is applying the same rule to the end of the file.
No clicks at the joins: Cutting silence out exactly at its edges and butting the two sides together leaves an audible click, because the waveform jumps from one value to another. On any recording with room tone it happens every time. Every join here gets a five-millisecond ramp: long enough to remove the step, short enough that the ramp itself is inaudible.
The audio itself is untouched: Only the timeline changes — no level shift, no filtering, no normalisation, nothing irreversible. What remains is the original samples in the original order, the five-millisecond ramps at the joins being the sole exception. The before and after durations, the percentage removed and the number of gaps are all reported, so you can check what happened in figures.

How to use it

Choose an audio file
Press the box to pick a file, or drop one onto it. Once it has loaded, the name, length and size are shown.
Decide what counts as silence
Three sliders: the threshold below which audio counts as silence, the shortest silence worth cutting, and the padding to keep at each end of a gap. The defaults are a reasonable first try.
Close the gaps, listen, save
Press "Close the gaps". You get the before and after lengths, the percentage removed, the number of gaps, and a strip showing where the cuts fell. Preview the result, then pick WAV or MP3 and press Download.

Choosing the settings

The threshold is your definition of quiet: 0 dB is the top of the scale and lower numbers mean quieter: −60 dB is close to true silence, −40 dB is about the noise floor of a quiet room. If only digital silence should count, start near −60 dB; for a recording with air conditioning, fan noise or tape hiss, −50 to −40 dB is the useful range. When detection misses, move this figure in 5 dB steps before touching anything else. Push it too high and quiet speech and breaths start being judged silent — a sudden jump in the number of gaps found is the warning sign. The measurement itself is an average over roughly twelve milliseconds rather than an instantaneous peak, so a single knock or a burst of equipment noise does not break a silence in two — at the cost of reacting less to very short sounds.
The minimum silence stops it over-tightening: Quiet stretches shorter than this are kept even when they fall below the threshold. Lower values close smaller gaps and shorten the file further, but below about 0.2 seconds the spaces between words and around punctuation start to go, and the same content comes back sounding hurried. For conversation and narration, 0.4 to 0.8 seconds is the comfortable range. If you only want to lose the waiting around at the start and between takes, set it to two seconds or more and the delivery is left completely alone.
Do not set the padding to zero: Cutting a silence exactly to its edges eats the attack of the next word and the tail of the last. Both are quiet, and both matter far more to comprehension than their level suggests, so the result reads as breathless rather than shorter. Keeping the default hundred milliseconds at each end is what makes the edit go unnoticed. If you need to remove more, shorten the minimum silence rather than the padding — it costs less in listenability.
It is not suited to instrumental music: The test is level alone, so it cannot tell a musically meaningful silence from empty space. A decaying note, the middle of a fade, a held pause between phrases — any of them can fall below the threshold and be treated as a gap. This is a tool for material where sound is either present or absent: talk, narration, lectures, interviews. On music, set the minimum silence to two seconds or more so only the spaces between pieces qualify, and always check the preview.

Questions

Is my file uploaded?: No. Reading, analysis and export all happen inside the browser. The result leaves only as a download you start yourself; neither the original nor the processed file is sent to a server. Unreleased recordings, internal meetings and interview material you are not free to share are all fine to run through it.
How long a file can it handle?: Up to about an hour is usually comfortable. The size ceiling is 200 MB, and anything larger is refused as you load it. Long files take a while to analyse, but progress is reported as a percentage rather than leaving the page apparently frozen. Note that the audio is expanded into memory to be examined, so on a modest device even a file under the limit can feel heavy — split it first if so.
The output file is bigger than the original: That happens when you save as WAV. WAV is uncompressed, so exporting a compressed source such as an MP3 grows the file even though the duration shrank; stereo WAV runs to roughly 10 MB a minute. Choose MP3 to keep the size comparable. If you plan to keep editing afterwards, though, staying with WAV avoids a second round of lossy compression.
Can it also raise the level or reduce noise?: No — this one only shortens the timeline. Level and tone belong to the mastering tool, separating vocals from backing to the stem separator, loudness measurement to the key & BPM analyser, key and tempo to the key & tempo changer, and format conversion to the file converter. The intended flow is to close the gaps here, save as WAV, and take that to whichever page does the next job. To make the recording in the first place, there is a voice recorder page.
Can I see where the cuts were made?: Yes. Below the figures, the whole file is drawn as a single strip: left edge is the start, right edge the end, and the shaded areas are what came out. Seeing whether the silence was all at the front or scattered throughout is what tells you which setting to adjust. The preview player underneath lets you hear the result directly.

Input

Output

Related tools

Highlights

How to use it

Choose an audio file

Decide what counts as silence

Close the gaps, listen, save

Choosing the settings

Questions