Video & Audio
Silence Remover (Cut Dead Air)
Automatically cut silent gaps from podcasts, voice notes and lectures — set the threshold and minimum pause, keep a natural breath, and save the tighter audio.
How to use
- 1Add a recording and set how quiet counts as silence and how long a pause must be.
- 2Preview the result (time saved is shown), then download.
Frequently asked questions
What threshold should I use?
−40 dB works for most clean voice recordings. Raise it (e.g. −35 dB) if background noise stops silences being detected.
Will it make speech sound rushed?
Keep 'Keep around cuts' at 0.1–0.2 s so words and breaths aren't clipped.
Can it remove 'ums'?
No — it removes silence only. Filler words need manual editing.
A worked example of the three settings working together
A 10-minute (600-second) interview recording with the default -40dB threshold and a 0.7-second minimum pause might report something like 'Removed 90.0s of silence (15%)' after processing — meaning every gap quieter than -40dB and lasting at least 0.7 seconds got cut, while shorter pauses (a natural breath between words) were left alone because they didn't clear the minimum-duration bar. The 'Keep around cuts' padding then adds back a small buffer (0.15s by default) on each side of every cut, so a sudden silence doesn't chop off the trailing consonant of the word right before it or the leading edge of the next word.
Why the threshold sometimes needs raising above the -40dB default
The threshold defines how quiet a moment has to be before it even counts as a candidate 'silence' — in a genuinely clean, quiet-room recording, true silence is very close to -60 or -70dB, and -40dB comfortably catches gaps between sentences. But a recording with a constant low hum (a fan, HVAC, room tone) might never actually dip below -35dB even during 'silent' pauses, which means a -40dB threshold would detect zero silence at all, not because the algorithm failed but because the recording's true noise floor sits above the threshold. Raising the threshold to -35 or -30dB in that case tells the tool to treat that noisier floor as the new baseline for 'silent enough to cut.'
Why this can't touch filler words, and what that says about how it actually works
The detection is based entirely on loudness and duration — it has no concept of words, meaning, or speech content at all, which is exactly why 'um' and 'uh' are untouched by design: a filler word is spoken at normal speaking volume, so acoustically it looks identical to any other word to a loudness-based detector. Only the genuine quiet gaps between words or sentences ever qualify as candidates for cutting. Removing filler words would require actual speech recognition to identify them as specific words first, a fundamentally different (and much more complex) kind of analysis than measuring amplitude against a threshold.
A pause has to stay quiet for whole 20-millisecond slices
Silence is checked in slices of 20 milliseconds, not on every individual sample. A slice counts as sound if the loudest moment in any channel reaches the threshold you set. At the default −40 dB, a peak under 1% of full volume is treated as quiet. A run of quiet slices is removed only when it lasts at least as long as Minimum pause. The default 0.7 seconds is 35 of those slices, so a shorter breath stays in the file. Afterward you are told how many seconds were removed, to one decimal, and what percent of the original length that was. Silence below only runs from −70 dB to −20 dB, so ordinary speaking volume is never treated as a gap you can cut.
Silence at the very start or end is removed with no cushion
Keep around cuts puts a little of the quiet back on each side of a cut so words are not clipped. The default 0.15 seconds comes back as about 0.16 seconds, eight of those 20-millisecond slices, on both sides of a gap long enough to remove. That cushion is skipped at the two ends of the file. If the recording opens with silence, the result starts on the first sound and the opening gap is discarded. If it ends on silence, the trailing quiet is discarded too. The cushion only applies to a pause that has sound on both sides. The slider runs from 0 to 0.5 seconds in steps of 0.05, and 0 means the cut sits flush against the sound.
Related tools
- Audio Noise Remover (Reduce Background Noise)
Remove hiss, hum and background noise from voice recordings, podcasts and videos' audio — adjustable strength, instant preview and download — processed privately in your browser.
- Audio Cutter & Trimmer (MP3, WAV, M4A)
Cut and trim audio online — drag to select the part you want, add fade in and fade out, and save as MP3 or WAV. Great for ringtones and clips.
- Mute Video (Remove Audio)
Remove the sound from a video instantly — no re-encoding, no quality loss — and download the silent MP4.
- Audio Normalizer (LUFS Loudness)
Normalize loudness to broadcast and streaming targets — −14 LUFS for Spotify/YouTube, −16 for podcasts, −23 for EBU R128 — measured to the ITU-R BS.1770 standard.
- Crop PDF (Trim Margins)
Trim the margins of PDF pages — set how much to cut from each side, on all pages or a selection — to remove white space or scanner borders.
- Remove Blank Pages from Scanned PDF
Find and delete the blank pages a duplex scanner leaves behind (empty backs, separator sheets), ignoring dust, edge shadows and grey paper. Check the marked pages, then save.