How to Cut the Dead Air Out of a Recording

A forty-two minute recording contains about six minutes of nothing — pauses while you thought, gaps while you found the next slide, the second and a half before every sentence. Cutting them by hand means finding 137 separate gaps and trimming each one.

The detection is mechanical. The judgement is in two settings and one decision nobody tells you about.

Set the threshold against your room, not a number

Silence detection works on level: anything below a threshold for longer than a minimum duration counts as a gap.

The threshold has to sit above your room tone and below your quietest speech. That gap is specific to your room and microphone, and copying a number from a tutorial is why detection either finds nothing or eats your quiet sentences.

Find it by looking at the level during a genuinely silent passage — the few seconds before you started talking. Set the threshold a few decibels above that.

If room tone and quiet speech are close together, the recording is too noisy for reliable detection and the fix is noise reduction first, or a better recording next time.

Padding is what stops it sounding clipped

This is the setting that makes the difference between an edit nobody notices and one that sounds like a machine did it.

Speech does not stop at the waveform. There is a breath before a sentence and a decaying tail after it, and both fall below the threshold. Cutting exactly where the level drops removes the breath and clips the last consonant, which the ear hears immediately even when it cannot name what is wrong.

Around 100 milliseconds either side restores it. More than 200 and the gaps start coming back.

Not every pause is dead air

The decision the tool cannot make.

A pause before an important point is doing work. So is the pause after a question, and the one where you let something land. Removing all of them produces a delivery with no space in it — technically tighter and harder to listen to.

The practical approach is to cut everything above about half a second, then watch back and reinstate the handful that were deliberate. That is a five-minute pass on a forty-minute recording, and it is the difference between a tight edit and a relentless one.

Keep the cut list, not just the output

Exporting a cut version and discarding the list is a decision you cannot revisit.

A cut list — the timestamps of each removed segment — lets you adjust the threshold, reinstate specific pauses, or start again from the original without re-detecting. It is a small text file and it is the difference between an edit you can revise and one you have to redo.

It also documents what was removed, which matters for anything where the recording is a record rather than a production. An interview edited for pace should have a list of what was cut, and for some purposes that list needs keeping alongside the original audio.

Most editors can import a cut list as markers, which makes reviewing the proposed edit far faster than listening through it.

Common mistakes to avoid

• Copying a threshold from a tutorial rather than measuring your own room tone.
• Setting the minimum silence too short, which removes natural pauses along with dead air.
• Cutting with no padding and wondering why the result sounds clipped.
• Applying the cut list without watching it back.
• Running silence detection on a recording with heavy background noise, where no threshold separates speech from room.

How to do it with Video Silence Auto-Cutter

The Video Silence Auto-Cutter detects the gaps in your browser and produces a cut list.

1. Load the video and check the level during a silent passage to set the threshold.
2. Set the minimum silence duration — half a second removes dead air while leaving natural pauses.
3. Add padding either side, starting at 100 ms.
4. Review the cut list and reinstate the pauses that were deliberate.

Other video tools are in the tools directory.

Frequently asked questions

What threshold should I use?

One a few decibels above your room tone, measured during a genuinely silent passage of your own recording. A number copied from elsewhere either detects nothing or cuts your quiet speech.

Why does the result sound choppy?

Almost always insufficient padding. Cutting exactly where the waveform drops removes the breath before a word and the tail of the previous one, which the ear registers as clipping.

Should every pause be removed?

No. Pauses before an important point are doing work. Cut the dead air automatically, then watch back and reinstate the deliberate ones.

Final thought

Measure your own room tone and add padding. Those two settings decide whether the edit is invisible or obviously automated.

Try the free Video Silence Auto-Cutter