Why audio-to-MIDI needs two different workflows

While building PureMIDI, I kept running into an overly broad label: “audio to MIDI.” A clean solo stem and a dense full mix are different transcription problems.

For a solo or isolated stem, note and onset accuracy matter most. For a full mix, instrument separation and cleanup time matter more. That is why PureMIDI uses Quick Melody for clean sources and Pro Song for editable multi-track drafts.

I’m curious how producers evaluate these tools: note accuracy, separation, or minutes of manual cleanup? If you have a useful 30–60 second test case, what kind of source should we benchmark next?

Tool: https://puremidi.com/audio-to-midi