How to clean up a voice recording in five steps

Most voice recordings, whether a podcast interview, a voice memo for a video, or a lecture, are ruined by the same few things: dead air at the start, a hum underneath, a speaker who drifts away from the mic, and a file ten times bigger than it needs to be. Each of those has a fix, and the order you apply them matters.

1. Trim the ends

Cut the fumbling before the first word and the "is it still recording?" after the last. The audio trimmer lets you drag the start and end on a waveform and export just the middle. Do this first so that the later steps are not analysing thirty seconds of room noise.

2. Remove the silences

Long pauses make a recording feel slow, and in an interview they add up to minutes. The silence remover detects gaps below a threshold and cuts them out, with a setting for how short a pause to leave so the speech still breathes. If you would rather cut manually at each pause, the silence finder lists the gaps with timestamps.

For a podcast where you want to pull one segment as a clip for social, the podcast clip cutter does the trim and the export in one step.

3. Deal with the noise

There are two kinds. A constant background hum, hiss or fan noise sits under everything and is best reduced with a filter: the audio filter has a high-pass setting that removes rumble below about 80 Hz, where voices have nothing useful anyway, and a low-pass for hiss above the range of speech.

The other kind is noise that is only audible in the gaps between words. A noise gate mutes the signal whenever it drops below a threshold, so the gaps go silent and the noise under the speech is masked by the speech itself. Set the threshold just above the noise floor and add a little release time so word endings are not clipped.

4. Even out the volume

Speakers lean in and out, and two people on one recording are never the same level. A dynamics compressor reduces the loud peaks so the whole recording can be brought up together; a ratio around 3:1 with a moderate threshold is a safe starting point for speech. Then use the audio normalizer to set the overall loudness to a target. Podcast platforms generally want around −16 LUFS for stereo and −19 LUFS for mono; the loudness meter measures where you have landed.

Do the compressor before the normalizer. Normalizing first just raises the peaks with everything else, and the compressor then pulls them back down.

5. Export at the right size

A WAV of an hour's speech is over 600 MB. Speech does not need that. The WAV to MP3 converter at 96 kbps is fine for mono voice and cuts the file to under 45 MB; the audio file compressor lets you pick a target size and does the conversion to match. Keep the trimmed, cleaned WAV as your master and export the compressed version from it whenever you need a new format.

The short version

Trim, cut silence, filter and gate the noise, compress then normalize, and export compressed. Each step runs in the browser on the file you already have, so there is no reason to send a raw recording anywhere.