Does MP3 compression hurt MIDI conversion?
Not in this clean, single-melody fixture: MP3 at 128, 192, and 320 kbps, AAC at 128 kbps, and Opus at 128 kbps each retained all 8 notes. That result applies to this source, not every compressed song.
Controlled codec check
| Encoded source | Median across 10 fixtures | Result |
|---|---|---|
| MP3 128 kbps | 8 retained; 0 missed; 0 extra | 8 / 8 notes retained |
| MP3 192 kbps | 8 retained; 0 missed; 0 extra | 8 / 8 notes retained |
| MP3 320 kbps | 8 retained; 0 missed; 0 extra | 8 / 8 notes retained |
| AAC 128 kbps | 8 retained; 0 missed; 0 extra | 8 / 8 notes retained |
| Opus 128 kbps | 8 retained; 0 missed; 0 extra | 8 / 8 notes retained |
What this result does and does not say
The fixture was an alternating C4/E4 melody: 8 no-gap notes, 80 ms per note, rendered first as mono 44.1 kHz 16-bit PCM WAV. It does not establish that platform compression is harmless for a full song, a chordal mix, or another source. It shows only that the tested encodes did not change retained, missed, or extra-event medians in this fixture.
Fixture conditions and limits
Each encoded source had 10 repetitions. The detector used Balanced
Melody / solo instrument mode, a 2048-sample analysis window, and a
1024-sample hop. A declared input note was valid at 50 ms or longer;
retained meant exactly one output overlapping at least 50% and within 50
cents; extra meant an unmatched output or second output for one input.
Detector SHA-256:
DDA185AC8FB6A179A4FCFA73FD4197ACC95B8B6F59B233F5D3E330F596309F9F.
These are controlled source checks, not a universal audio-compression or accuracy promise.