Picking a bitrate
Bitrate is how many bytes each second of audio is allowed to use. Higher means better sound and a larger file. The right choice depends almost entirely on whether you are compressing speech or music.
| Bitrate | Roughly per hour |
|---|---|
| 64 kbps mono | 29 MB |
| 96 kbps | 43 MB |
| 128 kbps | 58 MB |
| 192 kbps | 86 MB |
| 256 kbps | 115 MB |
For an interview, lecture, voice memo or podcast, 64 to 96 kbps in mono is genuinely enough and most listeners will not hear a difference. For music, start at 192.
Where the big savings are
If your file is a WAV, you are carrying uncompressed audio — roughly 10MB per minute in stereo. Converting to MP3 typically cuts that by 90 percent or more, which is why a voice recording that will not attach to an email suddenly does.
Mono is worth understanding properly, because it is widely misexplained. Converting to mono does not by itself shrink an MP3: at a fixed bitrate, a mono file and a stereo file are the same size. What changes is that all of that bitrate is now spent on one channel instead of two, so the audio sounds noticeably better.
The saving comes from combining the two. A single-microphone recording carries no real stereo information, so you can drop to mono and halve the bitrate and end up with a file half the size that sounds much the same. That is the move for interviews, lectures and voice memos.
Related things people need
If your recording is currently inside a video file, the video compressor can strip the picture out and leave you the audio, which is usually a fraction of the size.
To send several recordings together, bundle them into a ZIP.
Why audio files vary so enormously in size
Uncompressed audio is startlingly large. CD-quality stereo stores 44,100 samples per second per channel, each sample two bytes, which works out at roughly 10MB per minute. An hour-long interview recorded as WAV is around 600MB — larger than most feature films at streaming quality.
MP3 and similar formats reduce that by an order of magnitude using psychoacoustic modelling: analysing which sounds a listener can actually perceive and discarding the rest. A quiet sound immediately after a loud one is masked and can be removed. Frequencies above the range of human hearing serve no purpose. What remains is a file a tenth the size that most people cannot distinguish from the original.
Choosing a bitrate
Bitrate is how many bits of data each second of audio is allowed to use, and it is the main lever on both quality and size. The right value depends far more on content than on preference.
Speech occupies a narrow frequency range and compresses extremely well. At 64 kbps in mono, an interview or lecture sounds essentially identical to the original to almost every listener. Podcasts routinely publish at 64 to 96 kbps for exactly this reason.
Music is far more demanding, spanning the full frequency range with complex overlapping sounds. Below 128 kbps, cymbals and reverb start to sound washy. 192 kbps is comfortable for general listening, and 256 is where all but the most attentive listeners stop noticing anything.
Mono is the free win people miss
If your recording came from a single microphone — a phone voice memo, an interview, a meeting — both stereo channels contain effectively identical audio. Storing both doubles the file size to convey nothing.
Dropping to mono lets you halve the bitrate without a perceptible loss, because there was no stereo information to preserve. Mono at the same bitrate is the same size — it is the bitrate reduction that shrinks the file. For any spoken-word recording, this is usually the first thing to do, and it often solves the size problem on its own.
Only keep stereo for music, or for recordings where position genuinely matters — a concert, an ambient field recording, a conversation captured with two microphones deliberately spaced.
Re-encoding an MP3
Converting an MP3 to a lower bitrate works, but it is a lossy step applied to an already-lossy file, so quality degrades slightly more than the bitrate alone suggests. The encoder is compressing something from which detail has already been removed, and it cannot know which artefacts were part of the original sound.
In practice, going from 320 to 192 kbps is unnoticeable. From 192 to 128 is fine for speech. From 128 down to 64 is noticeable on music and acceptable on voice. Where you have the original uncompressed recording, encoding from that directly always gives a better result than re-encoding a compressed copy.
Questions
Can I compress an MP3 that is already an MP3?
Yes, by re-encoding it at a lower bitrate. Be aware this is a lossy step on top of a lossy file, so quality degrades slightly more than the bitrate alone suggests. Going from 320 to 128 kbps is usually fine; going from 128 to 64 is noticeable.
Is my audio uploaded?
No. Decoding and encoding both happen in your browser, on your device.
Why does a long file take a while?
Encoding is real computation and your device is doing all of it. Expect roughly ten to thirty seconds per hour of audio on a modern laptop, longer on a phone.