omnizart VS mt3

For open-source models, there are Omnizart [1] and magenta/mt3 [2].

I suppose these models are trained on western / pop music, so they may not work nicely on ethnic music. You could

[1] https://github.com/Music-and-Culture-Technology-Lab/omnizart

https://github.com/Music-and-Culture-Technology-Lab/omnizart

Honestly, if I'm feeling a bit lazy, I'll use MT3 and clean things up with quantizing + by ear. If I'm transcribing a monophonic melody, I'll typically find the key by ear and use that to help guide my efforts. Of course, key changes and modulations do throw a wrench in things, but doing either is typically enough to get something roughed in for my purposes. I don't typically need to transcribe something for my own songs, as I'm comfortable enough with composing that I can work off of chord notation or random chord shapes and play with inversions and added tones to get something that I feel sounds nice.

its ai generated midi

I used the notebook in this github repository

Direct MIDI transcription of Macabre Plaza using MT3.

Here's the original MIDI if anyone is curious. I used MT3 to speed up the transcription process. That is what always takes the longest when doing a musical remake of any sort.

Link to model repo

github: https://github.com/magenta/mt3