Compress audio

How to Compress

1. 1

   Add your audio files

   Any common format, up to 200 MB each and twenty at a time.

2. 2

   Pick a preset, or name a size

   Balanced suits music, Voice is for speech. Choosing a target size lets the tool work the bitrate out.

3. 3

   Click Compress and download

   The result is an MP3, with the bitrate that was actually used reported next to it.

What the presets do

Maximum quality encodes at 256 kbps and leaves the sample rate alone, the choice when the source is lossless and you want the file smaller without thinking about it again. Balanced is 128 kbps at 44,100 Hz and is where most music stops sounding obviously compressed on ordinary equipment. Small is 96 kbps, noticeably lighter and fine for background listening or sending over a slow connection. Voice is 48 kbps at 22,050 Hz and forces mono, because speech carries nothing above a few kilohertz worth protecting and a lecture recorded in stereo is two copies of the same voice. Each preset is a single decision, not a slider to fiddle with.

Naming a size instead of a bitrate

Most people arriving here do not want a bitrate. They want a file under twenty-five megabytes because that is what an email will take, or under sixteen because that is the upload limit somewhere. The target mode works backwards from the number you type. The first bitrate is calculated from the size and the duration rather than guessed, which usually lands on the right step immediately, and then at most three further attempts adjust it up or down along the standard bitrate ladder. Each attempt is a full pass over the file, so the mode is slower than the presets by design; that is the price of hitting a number rather than approximating it.

Why the result may come in slightly under

The tool stops as soon as the file lands between 88 and 100 per cent of your target, and it will never knowingly go over. That asymmetry is deliberate: exceeding the size you named defeats the point entirely, while landing a little under costs nothing. The bitrate ladder has fixed steps, 128, 112, 96 and so on, so there is not always a step that lands exactly in the band. When the two neighbouring steps straddle it, the tool keeps the larger of the ones that fit rather than the last one it tried.

When compression will not help

If the file is already an MP3 at or below the bitrate the preset would use, re-encoding it makes the sound worse without making the file meaningfully smaller. The tool detects that, leaves the file untouched and says so, rather than handing back a degraded copy. The same logic explains the occasional case where a compressed result comes out larger than the source: an Opus voice note at 24 kbps re-encoded at 128 kbps is bigger, and the page reports that rather than pretending it shrank.

Frequently asked questions

Guides

• MP3 vs WAV vs FLAC: Which One to Send, and Which to KeepMP3, WAV and FLAC solve three different problems, and picking the wrong one either wastes bandwidth or throws away quality for good.