MP3 to MIDI vs WAV to MIDI: Did the input format change the notes?
Published October 2, 2026 · Original site test
In one controlled 10.07-second C-major test, this site's MP3 and WAV inputs produced byte-identical MIDI files: 15 detected notes, two note tracks, and the same 266-byte export. The WAV was decoded from that very MP3, so this compares two ways of feeding the same recording to the converter. It does not show that WAV and MP3 are equally accurate for every recording.
How we made the comparison
We started with the public Basic Pitch C-major test MP3. Using FFmpeg, we decoded that file to a mono, 44.1 kHz, 16-bit PCM WAV. Repeating that decoding command produced a WAV with the same SHA-256 hash as the test WAV, confirming its provenance. This WAV does not restore information removed when the MP3 was encoded.
We then chose each file in turn on the live MIDI From MP3 converter in the same in-app browser. We left the default lower/upper pitch-range layout and 120 BPM MIDI export setting in place. Both inputs displayed 10.07 seconds and one audio channel. After each conversion, we saved the MIDI export and compared the actual files, not just the on-screen note counts.
What came out
| Observed result | MP3 input | Decoded WAV input |
|---|---|---|
| Detected notes | 15 | 15 |
| Pitch range | C4–C5 | C4–C5 |
| Note tracks | 2: lower 8, upper 7 | 2: lower 8, upper 7 |
| Export size | 266 bytes | 266 bytes |
| Export SHA-256 | 28875991e6010a9cab62d26053864c1f923ed32086fa1ebe31de01fa7eea98d2 for both files | |
The Standard MIDI File parser found one timing track and two note tracks, with 8 and 7 note-on events respectively. Lower and upper are pitch groups for editing; they are not two recognized instruments. For the note order and timing in the original MP3 export, see our first C-major test.
Does MP3 versus WAV matter?
For this particular MP3 and the WAV decoded from it, the answer was no: the converter exported identical MIDI bytes. That is a narrow result. A WAV made from an uncompressed original recording is a different source from a later MP3 copy; its note detection may differ. Different encoders, sample rates, channel mixes, noise, and instruments can also change the audio presented to the model. We did not test those cases here and do not assign an accuracy percentage.
When choosing a format for your own project, keep the cleanest original recording you have. If your only source is an MP3, converting it to WAV can help a browser or editor open it, but the change of file format alone cannot recover lost sound detail.
How can I check my own file?
- Keep one short source recording. If you make a WAV from its MP3, label it as a decoded copy rather than a new original.
- Convert the MP3 and WAV separately with the same tool version and export settings.
- Compare the pitch range and note count, then inspect individual note starts, lengths, and extra or missing notes in a MIDI editor. Matching counts alone do not prove matching MIDI.
- For an exact comparison, save both MIDI files and compare their SHA-256 hashes. Equal hashes establish identical saved bytes; unequal hashes tell you to inspect the note events.
Is MP3 to MIDI actually good?
It can provide an editable draft for a clean, isolated pitched line. In this C-major clip, the earlier test found the expected rise-and-fall note order. Neither that observation nor today's identical MP3/WAV exports proves reliable transcription of dense songs, drums, reverb, or quiet overlapping instruments. Listen against the source and correct the MIDI before using it in a project.