What we tested in the current converter
We generated a mono, 44,100 Hz source containing A4, C5, E5, and G5 for one second each. We saved the same samples as an uncompressed WAV and encoded them as a 128 kbps MP3. No video, performer, title, channel name, or copyrighted recording was used.
| Input | File size | Intended pitches found | MIDI result |
|---|---|---|---|
| WAV | 352,844 bytes | A4, C5, E5, G5 | 5 events: 4 long notes, 1 short fragment |
| 128 kbps MP3 | 64,783 bytes | A4, C5, E5, G5 | 6 events: 4 long notes, 2 short fragments |
The WAV decoded to exactly 4.000 seconds. The MP3 decoded to about 4.049 seconds and shifted the detected starts by roughly 0.023 seconds. Both outputs retained the four intended notes longer than 0.9 seconds. The extra MP3 event was a 0.059-second C5 fragment near a note boundary.
When WAV is the better input
Choose the original WAV when it is already available. WAV can preserve the source samples without lossy encoding, so it avoids adding compression artifacts before pitch analysis. In this test it also produced one fewer short fragment. That does not mean every WAV will beat every MP3: a noisy WAV can still be a worse source than a clean MP3.
When MP3 is a reasonable input
Use the existing MP3 when that is the file you legitimately have. The 128 kbps version in this test was less than one fifth the size of the WAV and still preserved all four intended pitches. Converting that MP3 to WAV later cannot restore audio information removed by the earlier lossy encode; it only places the decoded audio in a larger file.
What the converter actually accepts
The current local-file picker accepts MP3, WAV, and M4A. Each selected file is decoded by the browser and resampled to mono at 22,050 Hz before note analysis. M4A is accepted by the interface, but it was not part of this controlled comparison, so this page does not claim a measured M4A result.
A practical choice
- Keep the cleanest original you have. Avoid another lossy export before transcription.
- Use WAV directly when it is the source. There is no need to make an MP3 copy first.
- Use an existing MP3 directly. Rewrapping it as WAV does not recreate lost detail.
- Review short boundary notes. If they appear, try the minimum-note-length cleanup before downloading.
What this test does and does not prove
It verifies that the current converter processed both generated files and found the four intended pitches in this run. It does not establish an accuracy percentage, guarantee that WAV always creates fewer events, or compare every bitrate, codec, browser, or recording.
For simultaneous notes, see the controlled chord test. If a conversion produces nothing, use the no-notes checklist.