Can we create music with Artificial Intelligence? with François Pachet

[@rodneycollaco6645]

Burning question of the day, Can we upgrade or recreate exactly and save a low quality mp3 audio track (instrumental ,without vocals) to lossless CD quality with help of AI , if yes how ????

[@analogsamurai9576]

lookup OpenAI jukebox, the AI deconstructs audio masters, then the entire point is that the AI learns from the deconstructing an approximate way to reconstruct it(this can be tinkered with to reconstruct into a different genre or artist which it has learned from) and can generate novel songs with singing and such. For your question this is a good sign of being able to upscale a low quality audio(because that is exactly what the second reconstruction part of the jukebox is doing) at some point in the near future, given that the AI has been able to see something of greater quality. For example if we are trying to upscale a shoddy bach piano song, there is a relative amount of well defined bach piano music out there and the AI would easily be able to guess what notes go where, but if we are trying to upscale some noise music recorded on a phone at a concert it would have a hard time generating an accurate upscale. If we can upscale images, audio is not far behind but has a lot more variables.

[@closerfestival9078]

This has aged like a fine bossanova