Yes. I anticipate that the open source models will pave the way for that, just like we have in painting with stable diffusion.
Fundamentally, a song can be represented as a 2d image without any loss
Fundamentally, a song can be represented as a 2d image without any loss
- the song file stored in binary, printed out line by line
- the sheet music for the song, ie instructions for recreating it
In AI/ML world we're usually thinking about encoding into a series of high dimensional vectors, not sure off the bat how to represent that as a 2d image