A new deep learning AI algorithm developed by MIT CSAIL called PixelPlayer can isolate soundwaves of individual instruments in a video by recognizing the pixels from where the sound is coming thus enabling tuning of individual frequencies. MIT says the system is 'self-supervised' and doesn't require any human annotations.
https://www.notebookcheck.net/MIT-s-new-AI-system-can-tune-individual-musical-instruments-straight-out-of-a-concert-video.316620.0.html