Reduce Audio File Size
Make the file smaller. See exactly how small before you commit.
any size. There is no upload to limit
| Original file | · |
|---|---|
| Projected output | · |
| Saved | · |
Sample rate and channel count do not change the byte count of a lossy file, the bitrate alone decides that. What they change is how much the encoder has to describe with those bytes, which is why a mono 16 kHz voice recording sounds far better at 64 kbps than a stereo 44.1 kHz one does.
How to use it
- Drop the file that is too big onto the page. Its real size appears in the table below as soon as it is decoded.
- Lower the target bitrate until the projected size fits under whatever limit you are up against. The number updates as you change it.
- If the recording is speech, switch it to mono and drop the sample rate. Both let the same bitrate describe far less, so the result sounds better than the number suggests.
- Play it back before you commit. Aggressive settings are obvious on music long before they are on a voice.
- Click Download when the projected size and the sound are both acceptable.
Four things decide how big an audio file is
Duration, bitrate, sample rate and channel count. For a lossy format the arithmetic is blunt: bytes equal bitrate multiplied by seconds and divided by eight, plus a header. A 128 kbps file is 16 kilobytes per second of audio whether it contains a symphony or a cough. That makes bitrate the only control on this page that directly moves the number, and duration the only thing that moves it harder.
The other two work indirectly, and that is not a weakness. Sample rate sets the highest frequency the file can represent, half the rate, by Nyquist's theorem, so dropping from 44.1 kHz to 16 kHz tells the encoder it no longer has to describe anything above 8 kHz. Switching to mono halves the number of signals it has to describe at all. Neither saves a single byte, but both hand the encoder a much easier problem, and at a fixed bitrate an easier problem sounds better. That is why telephone-grade speech survives at rates that would destroy music.
Working backwards from a limit
Most people arrive here because something refused a file. Gmail stops at 25 MB per message, many corporate mail servers at 10 MB, and course platforms and forums often set their own smaller caps. Working backwards from the limit is simple arithmetic: a target of 10 MB over a 40-minute recording is 10 megabytes divided by 2400 seconds, about 4.2 kilobytes per second, which is 33 kbps. Too low for MP3, comfortable for Opus, and a sign that mono and a 16 kHz sample rate are not optional.
For reference points at ordinary settings: an hour of mono speech is about 29 MB at 64 kbps and 43 MB at 96 kbps. An hour of stereo music is roughly 58 MB at 128 kbps and 86 MB at 192 kbps. If those numbers do not get you under the limit, the answer is to make the recording shorter rather than to keep pushing the bitrate down. Cutting the dead air out of an interview routinely removes a fifth of its length without touching quality at all.
Two different things called compression
The word does double duty in audio and the two meanings have nothing to do with each other. This page performs data compression: fewer bytes on disk, at the cost of some fidelity. A dynamic range compressor is a mixing tool that reduces the gap between the loudest and quietest parts of a performance, making a recording sit at a more consistent level. It changes how the audio sounds and leaves the file size exactly where it was.
If you came here because a voice recording swings between shouting and whispering, the dynamics tool is what you want, and it is a separate page. If you came because an upload form rejected your file, you are in the right place. Doing both is a reasonable workflow. Level it first, then shrink it, because a compressed dynamic range is slightly cheaper to encode as well.