Remove dead air from a podcast, voice memo, or voiceover.

How it works

We scan the file from the start, looking for the first sample loud enough to cross the threshold you picked. Then we scan from the end backwards, looking for the last loud sample. Everything between is kept, plus a small padding so the audio doesn't start mid-syllable.

Picking the threshold

A quiet home studio recording has a noise floor around -55 to -65 dBFS. A phone recording in a normal room is closer to -45 dB. If the tool cuts too much, raise the threshold (less negative). If it leaves obvious silence in, lower it (more negative).

What it doesn't do

This trims leading and trailing silence, leaving everything between the first and last word exactly as recorded. If you also want the pauses inside the recording shortened, that is a different and much fiddlier operation — cutting mid-word, clicks at the joins, speech that ends up sounding rushed — and it now has its own tool: auto-cut silence. Use this one when you only need a clean entry and exit.

What this is good for

• Podcast editing — clip the silence before "Welcome to" and after the final outro tag.
• Voiceover takes — trim the dead space before and after the read so you can drop it straight into the timeline.
• Sermon / lecture / class recording — strip the 90 seconds of room tone before someone starts speaking.
• Voice memo cleanup — iPhone Voice Memos record from the moment you tap; this trims the fumble.
• Audiobook chapters — make every chapter start instantly.
• Auto-trim before transcription — Otter and Notta charge by audio length. Trimming dead air is free minutes.

What the threshold actually measures

"Silence" in a recording is almost never digital zero. A room has a noise floor — air conditioning, traffic, the hiss of the preamp itself — that typically sits somewhere between -60 and -40 dBFS. The threshold is the level below which audio is treated as nothing, so it has to sit above your room's noise floor and below your quietest real content.

If nothing is trimmed, the threshold is too low: the room tone is louder than you assumed, which is common with a laptop microphone or a phone recording in a live room. If the first word is clipped off, it is too high, and a quiet breath or a soft consonant has been mistaken for silence. Those two symptoms tell you which way to move it, and one adjustment usually settles it.

Why it only trims the ends

Removing the pauses between phrases is a much harder problem and a destructive one. Speech pauses carry meaning — the gap before a punchline, the breath that signals a new sentence — and an algorithm that deletes every gap over 400 ms produces a recording that sounds rushed and slightly inhuman, which is why the podcast editors that do this offer heavy per-gap controls rather than a single button.

Trimming the ends has no such ambiguity: the dead air before someone starts and after they finish is never meaningful. If you do need internal edits, cut the sections you want and rejoin them, which keeps the judgement with you.

Do this before you normalize, not after

Order matters more than people expect. Peak normalization is set by the single loudest sample in the file, and a door slam or a microphone bump in the dead air at the start counts. Trim first and that spike is gone, so the normalizer works from the actual content. Trim second and you have already normalized against a noise you were about to delete.

The same applies to transcription: trim, then convert. Whisper and its relatives occasionally hallucinate text out of long stretches of near-silence, and removing the dead lead-in makes that less likely as well as cheaper on the services that bill by the minute.

A padding note

A completely hard cut at the first sample of speech can sound abrupt, because a listener expects a beat of room tone before a voice. If the result feels like it starts too suddenly, that is usually the fix rather than the trim being wrong — leave a little of the original in with the cutter, or add a short fade with fade in and out.