Silence Remover
Finds every awkward pause and yanks it out.
or paste one from your clipboard
How to use it
- Drop the recording onto the page. It is scanned where it sits, so an hour-long interview costs you the decode and nothing else.
- Set the threshold. The level below which audio counts as silence. The summary underneath immediately tells you how many gaps were found at that setting.
- Raise the minimum gap length until the tool stops cutting the natural beats between words, and raise the padding until speech stops sounding clipped at the joins.
- When the summary says the right thing, pick a format and download. The audio that survives is untouched, only the gaps are gone.
What the tool calls silence
Silence in a recording is almost never silence. Even a treated room has a noise floor. The microphone's self-noise, the air handling, traffic through a wall. So the passage you think of as "nothing" still carries a signal. That is why the control is a threshold rather than a switch. The file is measured in 20-millisecond windows, each window's RMS level is compared against your number, and anything below it is a candidate for removal.
Twenty milliseconds is a deliberate choice. Shorter windows would trip on the moments a normal waveform passes through zero, which happens hundreds of times a second in ordinary speech. Longer windows would smear over a real breath and miss it. At 20 ms the measurement tracks the envelope of the sound rather than its individual cycles, which is what a listener hears anyway.
The second condition (minimum gap length) is what separates a pause from a rhythm. Ordinary speech contains dozens of sub-200-millisecond silences per minute, between words and inside stop consonants, and removing them makes a person sound like they are being played back wrong. Setting the minimum around 0.4 seconds ignores the rhythm of speech and catches only the pauses a listener would also notice.
Why padding matters more than it looks
When a stretch is removed, the cut is pulled inward by the padding amount at both ends, so some of the quiet stays. This is not politeness; it is a fix for a specific artefact. Plosives and fricatives (p, t, k, s) begin with a very low-energy approach before the burst, and a cut placed exactly where the level crosses the threshold slices into that approach. The result sounds like the speaker is being interrupted by their own recording.
Padding also protects the natural cadence of speech. Removing every millisecond of dead air produces a person who never breathes, which listeners find exhausting long before they can say why. Leaving 80 to 150 milliseconds either side keeps the impression of a pause while removing its length.
When not to use this
Anything where the silence is the content. Music uses rests structurally, and a tool that cannot tell a rest from dead air will flatten a piece. Recordings intended as evidence or as a record of a meeting should not have time removed at all. The edit is invisible in the result, which is exactly the problem.
And if the goal is a shorter file rather than a tighter one, this is an inefficient way to get there. Removing 90 seconds of silence from a 40-minute episode saves about 3 percent of the file. Dropping the bitrate or converting to a more efficient format saves far more, and does not change the edit.