That's impressive! The video got my attention immediately:

https://www.youtube.com/watch?v=FjkpVbbDtMY

> Grab a track from spotify, split into stem files, and then chop it up.

I'm guessing the "AI" plays a role in determining the tempo, vocal/instrument layers, where to chop up the beats.

The first few minutes of the video demonstrates how practical the application is, to drag a file/URL(?) in and start looping, mixing and matching.

It reminds me a bit of Ableton, with their automatic detection and marking of "beats" (amplitude peaks?).

I'd be curious to learn more about the technical aspects. For example, the readme says both Java and Python are required - but I didn't see any Python code in there. I suppose it's bundling an external library, probably for machine learning. I do see Java functions for training models - I wonder what datasets were used, what musical asepcts the pattern recognition works on.

The feature set is extensive, with synthesizers (like TB-303/808/909), sampler, live sessions, record/export - I can see it's a long-term project built up over years. Nice work!