What we tested
We generated three mono PCM WAV files containing the same isolated C4, E4, G4, and C5 sine tones. Each note lasted 0.75 seconds, with a 0.35-second gap. Every file used a 22,050 Hz sample rate; only the encoded bit depth changed: 8, 16, or 24 bits.
All three files went through the current local converter with its default controls. Each browser run returned HTTP 200, completed without console errors, and exported one musical MIDI track.
| PCM bit depth | WAV size | MIDI events | Detected pitches | MIDI size |
|---|---|---|---|---|
| 8-bit | 102,577 bytes | 6 | C4, E4, G4, C5 | 105 bytes |
| 16-bit | 205,110 bytes | 6 | C4, E4, G4, C5 | 105 bytes |
| 24-bit | 307,643 bytes | 7 | C4, E4, G4, C5 | 113 bytes |
What changed in the MIDI
The 8-bit result split C4 into two events and G4 into two events. The 16-bit result split C4 and E4. The 24-bit result split C4, E4, and G4, giving it one more event than either of the other files.
The main C4 event lasted 0.661 seconds at 8-bit and 0.697 seconds at both 16-bit and 24-bit. Main-note velocities also moved: C4 was 88 on the 0-127 MIDI scale at 8-bit and 83 in the other two results. These differences occurred even though the source tones, sample rate, timing, and peak level were otherwise the same.
Which bit depth should you use?
- Keep the best original you already have. Reducing bit depth discards level detail, while converting an 8-bit file to 24-bit later cannot restore that detail.
- Do not assume the largest file wins. The 24-bit input produced the most events in this test, but the extra event was a short fragment rather than an additional intended pitch.
- Review boundaries and velocities. All three outputs found the intended pitch classes, yet their event splits and velocity values differed.
- Treat conversion as a draft. Check short duplicate fragments before importing the MIDI into a larger arrangement.
What this test does not prove
It compares three exact PCM encodings of four synthetic notes. It does not rank every WAV encoder, measure real instruments or complete mixes, compare floating-point audio, or establish an ideal bit depth for every converter.
For a sample-rate comparison, see our 8,000, 22,050, and 44,100 Hz test. If the output contains unwanted fragments, use the extra-note cleanup guide.