The sizes, measured
MP3 bitrate is close to exactly predictable, because a constant-bitrate MP3 spends the same number of bits on every second regardless of content. The arithmetic is bitrate divided by eight, so:
| Bitrate | Per minute | Per hour |
|---|---|---|
| 128 kbps | 0.96 MB | 57.6 MB |
| 192 kbps | 1.44 MB | 86.4 MB |
| 256 kbps | 1.92 MB | 115.2 MB |
| 320 kbps | 2.40 MB | 144.0 MB |
Mono halves every figure. A one hour meeting recording at 128 kbps mono is about 29 MB.
When you can hear the difference
For speech, 128 kbps is transparent — there is nothing in a voice recording that a higher setting preserves. Podcasts are routinely published at 96 or 128 kbps mono for this reason.
For music the honest answer is that 192 kbps is where most people stop being able to tell in a normal listening situation, and 256 kbps is where nearly everyone does. 320 kbps exists because storage is cheap and certainty is worth something; it is not a placebo, but the material where it matters is narrow — dense cymbals, applause, some synthesised sounds.
The part that matters more than bitrate
If the audio you are converting is already compressed — and audio inside a video always is — you are re-encoding lossy material, which costs more quality than the bitrate choice does. Extracting to M4A copies the original audio instead of re-encoding it, and beats any MP3 setting for quality. Choose MP3 for compatibility, not for quality.