I love it when AI hallucinates. Not text models, that‘s no fun. But with audio models you can get some crazy sh#t like weird mono- and dialogues at the end of songs. Or completely f#cked up audio that not even Aphex Twin would be able to come up with 😃