Compress Audio

Re-encode audio at a lower bitrate, or down towards a size limit you have been handed.

Runs on your deviceAudioWorkingNo account needed

What this tool does

Compress Audio decodes your file in the browser, optionally reworks its channels and sample rate, then re-encodes it as a constant-bitrate MP3 with the LAME encoder this site serves alongside the page. Nothing is uploaded and nothing is queued on a server — the tab you already have open reads the file from disk and writes the result back. Whatever goes in, an MP3 comes out, named after the original with -compressed on the end.

Two routes lead to a smaller MP3. Choose a bitrate and you are picking a rung on a fixed ladder — 64, 96, 128, 160, 192, 256 or 320 kbps, with 128 as the default — and the size follows from it. Name a maximum file size instead and the bitrate becomes arithmetic rather than a search: a constant-bitrate MP3 is bitrate multiplied by duration, so the budget is divided by the length of the audio and the highest rung that fits underneath is taken. Because it is a ladder rather than a dial, you land under your figure rather than exactly on it.

Two further settings act before the encoder does, and it is worth knowing what they do not do. Mix down to mono folds the channels together; the sample rate list moves the file to 44.1, 32, 22.05 or 16 kHz. Neither makes the output smaller on its own, because a constant-bitrate MP3 weighs bitrate multiplied by duration whatever is handed to it — a mono 16 kHz file at 128 kbps comes out the same size as the stereo original at 128 kbps. What the two buy is room: half as much signal, or a narrower band of it, described by the same number of bits, so a spoken recording still holds together on a rung that would ruin music. The lower rung is where the saving actually happens. Rates the encoder does not support are snapped to the nearest one it does, and anything carrying more than two channels is folded to stereo first, because MP3 has nowhere to put the rest.

The limits are worth reading before you start. This is lossy encoding applied to whatever you feed it, so compressing an MP3 again decodes the damage the first encoder did and adds its own; keep the original if there is any chance you will need it. Asking for a higher bitrate than the source was written at recovers nothing and produces a larger file, and the summary says so when that happens. Picking 22.05 or 16 kHz drops the encoder into MPEG-2, where it stops at 160 kbps: pair one of those rates with 192, 256 or 320 and the file comes out at roughly 160 kbps while the details still report the rung you clicked. ID3 tags and cover art are not carried across, because the encoder writes audio frames and nothing else. And the whole file is decoded into memory as floating-point samples, so an hour-long recording will make a modest machine work for it.

How to use it

  1. Add your audio. MP3, WAV, M4A, AAC, FLAC and OGG are read, as far as your browser's own decoder goes. Queue several files and each is compressed on its own with the same settings.
  2. Choose which figure you are working to. A bitrate, when how it sounds is the thing that matters, or a maximum file size, when a form or a mail server has given you a number the file has to fit under.
  3. Set the figure. On the bitrate route, 128 kbps is the default and 64 kbps is the speech setting. On the size route, type the number, pick KB or MB, and the bitrate is worked out from the length of the audio.
  4. Fold to mono or lower the sample rate for speech. Neither shrinks a constant-bitrate file by itself. What they do is let a voice survive the 64 kbps rung, and it is that rung that shrinks it. Both are the wrong move for music.
  5. Download the result. The summary gives the old and new sizes with the percentage saved, alongside the bitrate, sample rate and channel count used.

Processing & privacy

Runs on your device. Your file is processed directly in this browser.

Common use cases

Getting a podcast episode under an upload ceiling

Hosting platforms and submission forms state a figure in megabytes. Name it on the size route and a bitrate is chosen that fits an episode of that length underneath it.

Emailing an interview recording

Dictation apps and field recorders write uncompressed or high-bitrate files that a mail server will simply refuse. The 64 kbps rung turns an hour of talk into about 28 MB, and folding it to mono is what keeps that rung listenable.

Lightening audio on a web page

A background loop or a pronunciation sample does not need 320 kbps. Moving it to 96 or 128 kbps cuts what every visitor downloads without anyone noticing the difference.

Fitting a library onto an old player

Devices with fixed storage and a USB socket take MP3 and little else. Re-encoding at 160 or 192 kbps roughly halves a 320 kbps collection and still sounds right on the speakers such a device is plugged into.

Questions

Does any of this audio leave my computer?
No. Decoding and MP3 encoding both happen in this browser tab. The encoder is a script this site serves, fetched the first time you compress something and reused from the browser cache after that; your audio itself is read from disk and written back without a request being made, so it is never transmitted.
Can I keep the original format?
No. Compress Audio always writes MP3, because MP3 is the only lossy encoder a browser can run without help from a server. Convert Audio handles format changes rather than size, but it is bound by the same constraint and cannot write AAC, FLAC or OGG either — only MP3 and WAV.
Will it hit the exact size I asked for?
It aims to come in under your figure, not to land on it. The bitrate is picked from a fixed set of seven, so the finished file usually sits some way below the number you typed — and where a rung falls almost exactly on your figure, MP3 frame padding can carry the result a little over, which the summary reports rather than hides. When even 64 kbps, the lowest rung, cannot fit the audio into the budget, the summary says the figure was not reached. Only a shorter recording helps from there: mono and a lower sample rate leave a constant-bitrate file the same size.
How much quality does the re-encode cost?
Something at every setting, and the loss is baked into the file that comes out. 192 kbps and above is hard to fault on ordinary listening, 128 kbps is the usual compromise, and 64 kbps is meant for speech and sounds thin on music. Nothing here restores what an earlier encoder threw away, so keep your source.
What happens to the title, artist and cover art?
They are not carried over. The encoder writes audio frames only, so the compressed file arrives with no ID3 tags and no artwork, and a player will fall back to the filename. Read what the original holds with Audio Metadata first if you intend to type them back in.
Why did my file get bigger instead of smaller?
Because the bitrate you chose is higher than the one the source was written at. Re-encoding a 128 kbps track at 320 kbps spends more bits describing audio that has already been through an encoder; the summary flags it when the output is no smaller than the input.
Which audio formats will it open?
Whatever your browser can decode: MP3, WAV, M4A and AAC, FLAC and OGG on most of them. WAV is read directly without the audio engine. WMA is supported by no browser at all, AIFF works only in Safari, and OGG and Opus vary between browsers — when a file cannot be read the error names the reason.
Can I compress a whole folder at once?
Add as many files as you like and each is compressed separately with the settings you chose, arriving back as its own MP3. There is no folder picker and no zip output, and a long list will take time on a modest machine, because every file is decoded in full before it is encoded.

Related tools