Automatically find long pauses, dead air and silent gaps in a recording, then shorten them to a natural length or cut them out completely. Compare the before and after waveforms, listen to the result and download it as MP3 or WAV, all without uploading your audio.
No file handy?
Anything quieter than this counts as silence. Raise it if background noise is not detected.
Shorter pauses, such as gaps between words, are left alone.
What to do with silences
Before silence that will be cut threshold
After
0:00.0
Silences found
#
Start
End
Length
Cut
Advertisement
Advertisement
How to use the Silence Remover
Open your recording: drop an MP3, WAV, M4A, OGG or FLAC file on the box, or click to choose one.
Check the threshold: it is set automatically from the background noise. The shaded parts of the Before waveform show the silences that were found.
Set the minimum length so only pauses longer than this are changed (0.5 s works well for speech).
Choose what to keep: shorten every pause to a few tenths of a second, type 0 to remove pauses completely, or trim only the start and end.
Listen and download: press Play result to check it, then Download without silence.
Recommended settings
Recording
Threshold
Minimum silence
Keep per pause
Podcast or interview
Auto (about −45 to −35 dB)
0.5 to 0.8 s
0.3 to 0.5 s
Voice-over or narration
Auto
0.4 s
0.2 to 0.3 s
Lecture or meeting recording
−35 dB (rooms are noisy)
1 s
0.5 s
Voice memo for transcription
Auto
0.5 s
0 to 0.15 s
Song or music track
−50 dB
0.5 s
Trim start and end only
How silence detection works
The tool splits the audio into 10 ms slices and measures each slice's RMS level in decibels relative to full scale: level (dBFS) = 20 × log10(RMS), where RMS is the square root of the average of the squared sample values. A full-scale sine wave is about −3 dBFS, normal speech is usually between −30 and −10 dBFS, and quiet room noise is often around −60 to −45 dBFS.
Slices below the threshold are silent. When silent slices last at least the minimum silence length, they form one silence. In Shorten mode, each silence longer than the keep value is cut down to that value, with half kept on each side so the speech before and after is not clipped. Silence at the very start and end keeps half the keep value as a lead-in and lead-out. Every cut gets a 5 ms fade on both sides, which prevents the clicks that appear when a waveform is cut mid-cycle.
Tips
Look at the dashed lines. They show the threshold on the waveform. Background noise should sit inside them and speech should reach well beyond them.
Use the list. Open Silences found and click a row to hear one second before that pause, which is a quick way to check the cuts.
Noisy recording? Run it through a noise reducer first, then remove silence, so pauses are easier to detect.
Frequently asked questions
What threshold should I use to detect silence?
Start with Detect automatically, which sets the threshold about 10 dB above the background noise. Clean studio recordings often work at −45 to −50 dB; recordings with fan, traffic or room noise may need −35 to −30 dB. If pauses are not detected, raise the threshold; if quiet words are being cut, lower it.
Will removing silence cut off quiet words?
Not if the settings fit the recording. Only parts that stay below the threshold for longer than the minimum silence length are touched, and by keeping a short piece of each pause (0.15 to 0.3 seconds), soft word endings and breaths are preserved. Listen to the result before downloading.
Should I remove pauses completely or shorten them?
For speech, shortening pauses to about 0.2 to 0.4 seconds sounds natural and keeps the rhythm of the speaker. Removing them completely (keep 0) gives the tightest result but can sound rushed. For music, use Trim start and end only.
Can it remove silence from the start and end of a song?
Yes. Choose Trim start and end only. Leading and trailing silence is cut, and pauses inside the track are left exactly as they are.
Is my recording uploaded?
No. The audio is decoded, analysed and cut in your browser with the Web Audio API, and MP3 files are encoded on your device. Nothing is sent to a server.
How does the tool decide what is silence?
It measures the loudness (RMS level in dBFS) of every 10 millisecond slice of the audio. A run of slices quieter than the threshold that lasts at least the minimum silence length counts as one silence. The loudest channel is used for stereo files.