What’s the currently best working audio to midi AI converter? I am trying that on a chiptune 
But melodyne fails…
This one looks fairly promising but I haven’t tried it. Uses the model from Spotify that is probably also running behind most “audio to midi” services, but it’s an open source and free vst plugin for all platforms.
oh thanks! I’ve just tried that, but it seems to detect one voice only, so you theoretically could convert a song to make it playable on a pc beep speaker thing 
I’m not sure that a good general-purpose polyphonic audio to MIDI converter even exists. It’s a hard problem. Other than Melodyne (and MIDI Guitar for guitar) I’m not even aware of a good special-purpose solution.
Edit: Have you tried the Samplab demo? I always kind of tune it out when I’m searching for this stuff due to my bias against companies that use “AI” as a selling point, but you know, I imagine current AI tech is a good fit for this task.
It transcribes as a single midi sequence but it does detect multiple voices and polyphony. I also gave it a quick test now and it did fairly well.
Well, had to manually edit a lot, that arp… 
I love this synth, so colorful yay
This so far seems to be the most accurate converter, single voice only:
https://midi.martinic.com/en/audio-to-midi
Tbh. I am kind of confused how the marketing claims about AI are vs. reality here… Still no algorithm seems to be able to extract polyphonic midi data accurately in 2025. Neither S1, Ableton, Cubase. And if the audio contains even a chiptune arp, all these tools will fail. Even the timing is wonky with all the tools.