I mean, we have AI generated art and now videos, and all the AI text anyone could want. How about music?
I was watching something or other with closed captions on, and among the cruft on the screen was something like [ominously suspenseful chords play] as some poor schnook was starting down the basement stairs. And it occurred to me that writing that sort of generic background ‘music’ might be just the sort of thing that AI could handle easily. I mean, yeah, the next Top 20 hit love song might be beyond it now, let alone the next great opera, but how hard is to to come up with a string of theramin-sounding chords in some minor key?
So, do you think that in a few months there will be a new Chat GPT when a movie director can type in “maniacally happy riotous party music, 24.5 seconds long, lots of cymbal crashes and saxophone frills” and it spits it out the final orchestrated version for you lickety split?
They are certainly not state-of-the-art, but you can try running prompts like that through AudioLDM2, Audiogen, and similar models and generate interesting samples that way. Some prompts seem to work better than others.
It’s been around for years, maybe even a decade. It’s nowhere near as good as high-quality human-composed music, of course, but it’s a lot cheaper, and good enough for ambient things like video game soundtracks.
Damn. While it’s not exactly the most creative lyrically, I’m impressed how suno.ai does. This was from a prompt about the singularity and the impending AI enslavement. Took it like 10 seconds to form. Holy shit.