I'm digitizing old bootlegs I recorded to audiocasette and just discovered that some old soundboard recordings split guitars to R, drums to L and vocals mixed across the two. That was...useful and kind. Thank you sound guy 27 years ago.
@jesse Have you looked at ML stem extraction? Neural nets are trained to separate an audio file into its independent instruments. For many mainstream pop recordings, it's uncannily good.
Spleeter is the most popular, but demucs gives better results in my experience.