How Big Is an Audio File?
We encoded the same 60 seconds 32 different ways and measured every byte. Some formats hit their target exactly. One missed by half.
Short answer: for MP3 at a constant bitrate the size is exact arithmetic — kbps × 0.0075 = MB per minute, so 128 kbps is 0.961MB per minute and 57.7MB per hour. Uncompressed WAV is equally exact. Everything else, including Opus, Vorbis and FLAC, depends on what the audio actually contains, in some cases by a factor of six.
Most pages that answer this question quote a single number per format. That is right for two of them and misleading for the rest. Below are our own measurements, with the method stated so you can reproduce or challenge them.
Method
Three 60-second 16-bit 44.1 kHz stereo WAV sources were generated with FFmpeg 6.1.1, each exactly 10,584,078 bytes, chosen to bracket the range of how compressible audio can be:
- Pink noise — broadband and unpredictable. The hardest case for any compressor, and a realistic floor.
- Harmonic stack — four steady sine partials. Highly predictable, and an unrealistic best case.
- Speech-like — pink noise band-limited to 300–3400 Hz with amplitude movement and pauses, roughly the shape of a voice recording.
Each source was encoded into 32 format and setting combinations and the output measured in bytes. Real music sits between the noise and tone extremes, so for the content-dependent formats treat the two as bounds rather than predictions. For the constant-bitrate formats content is irrelevant, which the measurements demonstrate. Throughout, MB means 1,000,000 bytes.
MP3 at a constant bitrate: exact, every time
All three sources produced byte-identical files at every bitrate — a spread of 0.0%. The size of a CBR MP3 is a property of the setting, not the music.
| Bitrate | Per minute · per hour · minutes that fit in 100MB |
|---|---|
| 64 kbps | 0.480MB · 28.8MB · 208 min |
| 96 kbps | 0.721MB · 43.3MB · 138 min |
| 128 kbps | 0.961MB · 57.7MB · 104 min |
| 160 kbps | 1.201MB · 72.1MB · 83 min |
| 192 kbps | 1.441MB · 86.5MB · 69 min |
| 256 kbps | 1.922MB · 115.3MB · 52 min |
| 320 kbps | 2.402MB · 144.1MB · 41 min |
The arithmetic is kbps × 60 ÷ 8 ÷ 1000 megabytes per minute. Measurement exceeded it by about 0.1% — a fixed header and frame padding, not a variable.
The last column is the practical one: our tools accept 100MB per upload, so a 320 kbps file can be about 41 minutes long, and a 128 kbps file about an hour and three quarters.
The finding most people get wrong: mono does not shrink a CBR MP3
We encoded 64, 96 and 128 kbps in both stereo and mono. The mono files were exactly the same size as the stereo ones — 0.480MB, 0.721MB and 0.961MB respectively, identical to the byte.
This surprises people, and the reason is simple once stated: a constant bitrate means a constant number of bits per second, whatever you put in them. Switching to mono does not reduce the budget, it spends the whole budget on one channel instead of two. You get better quality per channel at the same size, which is why mono is a good choice for speech — but if your goal is a smaller file, lower the bitrate. Changing channels alone achieves nothing.
Where encoders ignore your bitrate
"Bitrate" means different things to different encoders. Spread below is the difference between the largest and smallest of the three sources at the same setting.
| MP3 CBR | Spread 0.0% — the setting is the size |
|---|---|
| AAC | Spread 0.3–1.3% up to 192 kbps; 11.8% at 256 kbps, where it declines to spend the budget on simple material |
| Opus | Spread 81–100%. Asked for 32 kbps it delivered 0.178–0.357MB/min against a nominal 0.240MB |
| Vorbis (q) | Spread 225–594%. At q8: 1.888MB for noise, 0.272MB for tones |
| MP3 VBR (V0–V5) | Spread 68–152%. V2 ranged 0.592–1.346MB/min |
None of this is a defect. Opus and Vorbis are designed to spend bits where they are needed and save them where they are not, which is why they sound good at low bitrates. It does mean that if you need a predictable file size, constant-bitrate MP3 or AAC is the only reliable choice, and if you need to hit a hard limit, MP3 CBR is the one you can calculate in advance.
Lossless: the content decides, not the format
All figures per minute, against the identical 10.584MB WAV source.
| WAV 16-bit 44.1k stereo | 10.584MB — the baseline |
|---|---|
| FLAC, noise | 8.365MB — only 21% saved |
| FLAC, speech-like | 2.779MB — 74% saved |
| FLAC, tonal | 1.878MB — 82% saved |
| ALAC, same three | 8.454MB · 2.959MB · 3.090MB — consistently behind FLAC |
The familiar claim that "FLAC is about half the size of WAV" is a reasonable average for real music and useless as a guarantee. A dense, loud, noisy master lands near the 21% end; a sparse acoustic recording lands near the other. Nothing about the format tells you which — only the audio does.
ALAC was larger than FLAC on all three sources, by 1% on noise and 65% on tonal material. If Apple compatibility is not a requirement, FLAC is the better container.
FLAC compression level is nearly pointless
Level 8 against the default level 5 gave a byte-identical file for noise and 4% smaller for tonal material, for considerably more encoding time. The default is the right setting.
Uncompressed audio: pure arithmetic
sample rate × (bit depth ÷ 8) × channels × seconds. Every measurement matched exactly.
| 16-bit 44.1 kHz stereo | 10.584MB/min — CD quality. 9.4 min fits in 100MB |
|---|---|
| 16-bit 44.1 kHz mono | 5.292MB/min — here mono does halve it |
| 24-bit 48 kHz stereo | 17.280MB/min — studio capture |
| 16-bit 22.05 kHz mono | 2.646MB/min — voice memo territory |
Note the contrast with MP3: for uncompressed audio, halving the channels really does halve the file, because there is no bitrate budget — each sample is simply written down.
What to take from this
- Need a predictable size? CBR MP3. Multiply kbps by 0.0075 for MB per minute and you are done.
- Need to fit an upload limit? Divide the limit by that number for your maximum duration, or trim the file.
- Want a smaller file? Lower the bitrate. Switching to mono alone does nothing at CBR.
- Want the smallest file at listenable quality? Opus, accepting that you cannot predict the size.
- Archiving? FLAC at the default level. Expect anywhere from 20% to 80% saved and do not plan around a specific figure.
- Editing? WAV, and budget 10.6MB per minute at CD quality.
Related guides
- MP3 bitrate guide — what each level sounds like, as opposed to what it weighs
- Reducing MP3 file size — getting under a limit in practice
- Converting audio formats — choosing a container
- Does re-saving lose quality? — we measured that too, and MP3 behaves worse than you would hope
Summary
Two formats let you know the answer before you encode: constant-bitrate MP3 and uncompressed PCM. Both are simple arithmetic and both were confirmed to the byte. Everything else trades predictability for efficiency, and the variation is not small — Vorbis at a fixed quality setting produced files six times apart depending only on what it was fed. Quote a single number for those formats and you will be wrong most of the time.