I've been using machine learning to separate the music from the soundtrack of the 2001 film Osmosis Jones. Using selective muting and many, many spectral edits, I've extracted around 20 songs so far. There will be artifacts, but it's the best that can I can do with the sources and tech I have.

There seems to be a lot of AI audio tools; anyone try them for audio isolation? My usecase is separating a song from a 5.1 movie mix, there's a great remix of Hot Blooded I wanna preserve