@
0 reply
0 recast
0 reaction
Giuliano Giacaglia
@giu
This is very good
1 reply
0 recast
1 reaction
3LAU
@3lau
Wild, the principle here is really smart. Don’t have the model build the sound from scratch but have it write the song w/ midi and a library of one-shot audio samples w/ solid metadata mapping. Half human created (the samples, not loops) but the song is quite literally “written” by the ML - higher qual output!
2 replies
0 recast
1 reaction
David Moon
@davidmoon
I have to ask you a question. 1. How long did it take you to train this? 2. Would it be possible to train an AI in real time to reflect venue preferences? DJs get to control music at a rave. Being able to A/B test new or old music (remix) with accurate data collection would be huge…
0 reply
0 recast
0 reaction