Silence remover
Cut the dead air out of a recording without cutting the breaths that make it sound human. You see every gap it plans to shorten before anything is removed, and the file never leaves your device.
Speech is written as a mono MP3, small enough to upload anywhere - about half a megabyte per minute. Pick WAV instead if the file is going into an editor and you want the audio untouched.
Need the words as well as the audio? Transcribe it - 15 minutes free →Removing silence is not the same as deleting it
The reason automatic silence removal so often sounds wrong is that it treats a pause as waste. It is not: a pause is where a sentence ends, where a thought lands, where a listener catches up. Splice two words together with nothing between them and the ear hears a glitch rather than a tighter edit. So this tool shortens gaps rather than deleting them - it leaves a slice of the original room tone in place, and keeps a margin either side of the speech so no word loses its first or last moment.
The waveform shows exactly what will go before you commit to anything, and the preview plays the result without touching the file. For the other kind of tightening - the ums, the sos, the you-knows - the filler word counter shows what you actually say and how often. Once the episode is cut, a transcript gives you the show notes, chapters and subtitles in one pass, and this is the workflow for that. Working from video? Pull the audio out first.
Silence remover - FAQ
Is my recording uploaded?
No. The file is decoded and processed inside this tab using your browser's own audio engine. Nothing is sent anywhere, and the page keeps working with the network switched off.
Why is the download a WAV and not an MP3?
A browser can write a WAV file directly, in a few lines of code. Writing an MP3 or an AAC means shipping a full encoder to your device, which would make this page many times heavier for something your editor already does well. Import the WAV and export whatever format you need.
How big is the WAV going to be?
Roughly 10 MB per minute of stereo audio at 44.1 kHz, regardless of how small the original was. A trimmed one-hour episode lands around 500 MB. That is normal for uncompressed audio and it compresses straight back down when you re-encode it.
What threshold should I use?
Start at -40 dB, which suits a decently quiet room. If speech is being cut, raise the number towards -30. If nothing is being found, your noise floor is high - lower it towards -50 and consider whether the recording needs noise reduction first.
Why is there a "leave behind" setting?
Because total removal sounds broken. Cutting a two-second gap to zero makes two sentences collide; cutting it to a quarter of a second keeps the rhythm and the room tone while still saving you most of the time. That is why it defaults to 250 ms rather than nothing.
How long a file can it handle?
The whole file is decoded into memory, so it depends on your device. An hour of audio is comfortable on any modern laptop; three hours is the hard limit here, and phones will struggle well before that. For a long recording, split it first.
Does it work on video files?
It reads the audio track out of an MP4 or MOV and gives you back audio only. To re-cut the video itself you would need the timings in an editor - this tool is for the audio.
Will it remove background noise too?
No. It only shortens stretches that are already quiet - it does not change how anything sounds. Noise reduction is a different job and belongs in an editor, before this step.