Why WAV is a good input format for transcription
A normal PCM WAV does not use the kind of psychoacoustic compression found in MP3 or AAC. That does not magically make note recognition perfect, but it means the transcription model receives the waveform without additional lossy encoding artifacts. Clean attacks, quiet overtones, and note decays can be easier to inspect when they have not been altered by a low-bitrate encoder. If you recorded an instrument yourself or can export a stem from a DAW, a WAV copy is a sensible source for an audio-to-MIDI pass.
Bit depth and sample rate are less important than the musical separation in the recording. A pristine 24-bit/96 kHz full mix can still be much harder to transcribe than a simple 16-bit/44.1 kHz solo instrument. The model ultimately needs pitch and timing clarity. If your WAV contains a whole song, consider exporting the melody, bass, piano, or other target instrument as a separate WAV stem before conversion. That single decision often improves the usefulness of the MIDI more than increasing technical audio resolution.