fixmint
EN

How to cut the dead air from a recording and keep it natural

By · · 4 min read

Shorten the long pauses instead of deleting every one. A recording sounds natural when each gap keeps about the length a speaker would leave. It sounds chopped when every gap is removed and sentences run into each other. Then clean the hiss and hum, and last of all even out the loudness. Work in that order, because each step changes what the next one measures.

Remove silenceLong pauses in my recordingDo it free

Where the dead air comes from

A speaker stops to think, waits for a question or takes a breath. During those gaps the microphone keeps picking up the room: a fan, a chair, traffic outside. A long lecture or an unedited interview can hold many minutes of that. Cutting it shortens the file without changing a single word.

A silence finder measures the level in short slices of time. Any slice that stays below a chosen threshold for longer than a set minimum counts as a pause. The threshold has to sit above the room noise and below the quietest voice. Set it too high and soft words get cut. Set it too low and the room noise counts as speech, so nothing is removed.

Shorten the pauses, don't delete them

Speech needs its pauses. A full stop, a breath and a change of thought each take a moment, and listeners rely on that moment. Deleting every gap makes the speaker sound rushed and hard to follow. A short, fixed gap in each place keeps the rhythm.

The remove the silence tool can cut pauses out, or shorten each one instead. For speech, shortening is the safer choice. Two rules keep it clean:

  • Leave a long pause alone when it carries meaning, such as a beat before a punchline or the silence after a hard question.
  • Give each cut a little room. A cut right at the edge of a word clips its first or last sound, and the word sounds broken.

After a cut, play the edited part once at normal speed. A cut that sounds wrong usually sounds wrong on the first listen.

Clean the hiss and hum

Three kinds of noise cause most trouble in voice recordings:

NoiseWhat it sounds likeWhere it comes from
HissA steady shhh, like a radio between stationsThe microphone, its amplifier and a noisy room
HumA low, steady toneMains electricity: 50 Hz (cycles per second) in most countries, 60 Hz in North America
RumbleDeep, low noiseAir conditioners, traffic, handling the microphone

Each sits in a different part of the sound range, so no single setting fixes all three. Run the voice cleanup once, then listen to a quiet pause and a loud sentence. Stop when the noise is gone. If the voice starts to sound watery or metallic, the cleanup has taken too much of the voice along with the noise.

Even out the loudness

Two parts of a recording can be equally loud and still sound different. Loudness has two common measures:

  • Peak is the single loudest moment. One loud laugh gives a high peak, even if the speech around it is quiet.
  • LUFS (Loudness Units relative to Full Scale) measures the average loudness people actually hear across the whole file. It weights the frequencies the way human hearing does.

For a podcast or a lecture series, pick one target and use it for every episode. −16 LUFS is a common target for podcasts. Match the loudness brings each file to that one level, so a quiet episode and a loud one sound alike. It cannot repair a clipped recording. Clipping flattens the top of the wave when the input is overloaded, and the lost peaks are gone for good.

Loudness is the last step because the earlier ones change the level. Cutting and cleaning both change how loud the file measures, so measure after you finish them.

Edge cases

  • Quiet recordings with hiss. Raising a quiet file raises its hiss with it. Clean first, then level. That is one more reason the order matters.
  • Hiss too strong for the silence finder. If the noise is so loud that the pauses never drop below the threshold, clean the audio first, then cut the pauses from the cleaned copy.
  • Two people on one track. A pause between speakers often holds a breath or a laugh. Listen to the cut once with both voices before you accept it.
  • Long files. Everything runs in this tab, so an hour of stereo sound needs a lot of memory. Very long files work best on a computer.
  • The output is WAV. WAV keeps the full quality and opens on every editor and phone. If you need a smaller file, convert it to MP3 afterwards.
  • Keep the original. The tools return a new file and leave the recording you started with untouched. Keep it until the episode is published.

Comments

    No comments yet. Be the first.

    Your rating (optional)

    Your email is never shown or kept.

    Cover of Digital Workflow ToolkitDigital Workflow Toolkit

    Recipes, templates and patterns for chaining small file jobs (convert, rename, compress, merge) into workflows that run the same way every time.

    $19.00Get it on Gumroad