Remove Silence from Audio

Dead air is the easiest thing to cut from a recording and the most tedious to cut by hand. This tool finds every silent stretch, shows you exactly which ones it would remove, and lets you keep as much of each pause as the material needs.

A pause that felt natural while recording reads as a hole on playback. Removing the long ones makes an episode shorter without making it faster, keeps a lecture recording listenable, and turns a rambling voice memo into something you would send to somebody. The work itself is mechanical — find the quiet parts, delete them, join the ends — which is exactly the kind of work worth handing over, provided you can see what is being deleted.

• Shorter recordings without speaking any faster — the content is untouched, only the gaps go.
• Nothing to install and nothing to upload; the whole job runs on your device.
• Every cut is visible before it happens, so a threshold that would swallow quiet speech is obvious rather than discovered afterwards.
• MP3 in and MP3 out keeps the original audio exactly, because the parts you keep are copied rather than re-encoded.
• Up to 25 recordings in one pass with Pro, returned as a single ZIP.

This is the part that makes the rest usable. As soon as a file lands it is analysed once, and every stretch that falls below the threshold is shaded on the waveform. Beside it sits the count, the total time about to be removed, and the length the file will end up. Move the threshold and the shading moves with it. Nothing has been written at this point — you are looking at a proposal, not a result.

• The picture cannot lie to you. The shading is drawn from the same list of regions the cut is made from, not from a second calculation that could drift away from it.
• The threshold comes from your recording. A fixed default has no relationship to the file in front of it — a treated room sits far below the usual -50 dB and a cafe recording sits above it. The suggestion is measured from this file's own noise floor.
• Only two controls cost a re-analysis. The threshold and the minimum length need a fresh pass over the audio. The pause to leave in, the mode, the output format and the taper all recalculate instantly from the analysis already done.

Two passes, and the first one is the one you interact with. The first reads the audio and notes every stretch that stays below your threshold for at least your minimum length; that is what the cut map draws. The second builds the finished file from the parts that survive, joining them in order.

• Detection. The recording is read once. Every region below the threshold for long enough is recorded, along with the noise floor used to suggest a threshold in the first place.
• Planning. Your settings turn that list into the list of regions to keep — shortening each pause rather than deleting it, honouring the mode you chose, and working in whole samples so the finished length is known exactly before anything runs.
• Writing. Where the format allows it the kept regions are copied across untouched. Otherwise they are re-encoded into the format you asked for, with a short taper at each join. The finished length is checked against the planned one before the file is handed back.

Powered by optimised audio processing algorithms.

Speech has rhythm, and the rhythm lives in the pauses. Deleting them entirely produces a recording that is technically shorter and unpleasant to listen to — every breath gone, every sentence running into the next. Leaving 100 to 250 milliseconds of each pause behind keeps the shape of ordinary conversation while still removing the dead air, and it is the single setting that most affects whether the result sounds edited or sounds broken.

• Speech and interviews: 0.15 to 0.25 s. Enough to breathe, short enough that the tightening is obvious.
• A tight promotional read: 0.05 to 0.1 s, where urgency matters more than naturalness.
• Lectures and audiobooks: 0.3 s or more. A listener following an argument needs the beats between sentences.
• The bound is not arbitrary. You cannot leave more of a pause than the length that made it count as a pause, so this control's maximum follows the minimum silence length as you change it.

Most tools in this category decode your file, cut it, and encode it again — which loses a generation of quality every time, for an operation that removes audio rather than changing it. This one does not have to. When the input and the output are both MP3, and nothing you have asked for changes the samples, the parts you keep are copied across exactly as they were stored, with no decoding and no encoding in between. It is both faster and mathematically identical to the original.

• When the copy applies: MP3 in, MP3 out, with no change of sample rate and smooth joins turned off. The panel says so when it happens.
• Lossless formats stay lossless: WAV and FLAC keep every sample whether or not the copy path is available, because neither format throws anything away.
• What forces a re-encode: Changing format, changing sample rate or bit depth, or turning on smooth joins — a taper has to touch the samples, so it cannot be a copy.
• One honest limitation: A copied cut lands on the nearest frame boundary, so a join can sit up to about 26 milliseconds from where it was planned. Those boundaries are inside silence, so it is inaudible — but it is why the taper option needs a re-encode.

• Take the suggestion first. It is measured from your recording rather than guessed, and it is right far more often than -50 dB is. Adjust from there if the cut map disagrees with you.
• If quiet speech is being cut, lower the threshold. A more negative number is more selective — it treats less as silence. If too much dead air survives, raise it toward -40 dB.
• Use the minimum length to protect natural pauses. 0.4 to 0.7 s leaves the beats between sentences alone while still catching long gaps. Below 0.2 s the tool will cut between words.
• Leave the joins smooth. The taper costs nothing in length and removes any chance of a click at a cut. Turn it off only when you want the byte-for-byte MP3 copy.
• For music, go quieter and longer. Around -65 dB with a minimum of one to two seconds, so quiet passages and fades are not mistaken for silence.

• Podcasts: Cutting the dead air out of a conversational episode, which is where the biggest savings are and where leaving part of each pause matters most.
• Interviews and research recordings: Tightening the gaps between question and answer without losing the pauses that carry meaning.
• Voice-over and narration: Removing the run-in and run-out around each take, and the retake gaps in between.
• Lectures and meetings: Shortening a long recording so it can actually be reviewed, while keeping every word of it.
• Music and demos: Trimming the silence at the head and tail of a bounce, using the Start and end only mode so nothing inside the track is touched.

• Audio cutter
• Audio normalizer
• Noise reduction
• Audio compressor
• Audio converter