I have just lend my account to ChatGPT for 1 post:
"Hey there fellow Elektronauts! 
I wanted to introduce you to a fantastic track called “Melodic Techno” crafted by one of our talented AI companions, MusicLM. This track delves deep into the heart of melodic techno, with hypnotic rhythms, lush synthscapes, and groovy beats that’ll transport you to the cutting edge of electronic music."
I’m assuming this is a joke? I can’t hear any techno and it’s far from cutting edge. Sounds like generic commercial dance music of sorts - not that there’s anything wrong with that at all!
I did use chatgpt to make a rap about cheese , it did a Gouda job too.
I haven’t turned it into a horrendous modern trap auto tuned ballad yet though … It couldn’t be much worse than the human equivalent.
well, it´s all actual AI generated everything… but who knows which sounds playlist are these models trained on.
Can AI do this?
As crap as one would expect.
I find it really fascinating. I don’t think generic is the right way to describe the music (even though it is generated). Sounds more like a talented but very inexperienced kid to me.
Was the whole piece generated as one piece or is it compiled by several generated files?
I mean, we are on a forum hailing the creation of several synths and drum machines.
Don’t forget the scepticism surrounding both of those things when they first appeared.
Can’t tell if you’re hating on Mr. Fator, but that Dancing in the street-video is my favourite piece of contemporary art.
Got some fun stuff from https://www.stableaudio.com/ recently.
Not into making generic tosh with it (there’s plenty of that in the world already) but trying to push it to make intriguing sample fodder is fun.
one single piece, the IA keeps changing the patterns. you can hear all “productions” here:
nice. How did you prompt that bit?
prompt: ‘Ryoji Ikeda and Bjork together in 1990 Berlin with lots of c*caine’
interesting results.
Much better results at things that are full of common places:
This is just mind blowing! It truly sounds like Pac and the other rappers on the album. Some laughs here and there sounds off, but overall it sounds like Pac.
so i’m late to the party as usual after a friend introduced me to riffusion, hadn’t seen it since it was spectrogram based, i’ve done the usual rabbit hole prompt explosion, and it’s fun! it’s not great, and there’s beauty in when it goes wrong. it seems to have been trained on mastered audio (maybe sample collections etc?) so the output can have great sounding results (albeit mp3 currently)
the audio i’ve shared here i put a bit in current sounds to see if any innocent reactions. none. 
but sharing this cos i’m left feeling exactly the same way i do about image generators.
context for the audio is about 15 “remixes” that were cherry picked, and human ordered into something cohesive and transition “cleanly” between each clip, then processed to lift the low mp3 quality with extra high end. the sub is pretty much what rendered, but squashed a bit. it’s a bit much for sure.
and it’s cool! i enjoy the sound, there’s fresh elements to its structure and textural folding, but am left feeling uncomfortable and sad. (standard existential overthink from me).
it can do more generic stuff, and some clear pop formulas are clearly interwoven into the model being used especially if you use lyrics. but there’s lots beyond that!
it’s given me insight into how people like patten have done their recent “second ever total ai album” stuff.
if this is the quality now very soon many people will be pressing enter and creating music with a pre ordained palette and human influenced decisions/tags etc. and at a quality for idents and other commercial uses.
are we headed towards a model where all recorded music ever will be utilised for a result? that’s cool, but also worrying.
i hope for some boffins to create a system where you can train your own audio (i know this is probably already possible, but as a non tech wizard i’m waiting for the path of least resistance).
but ultimately as a sampling source this is great, for hits, pads, textures, what ever really.
and think i’m after some reassurance/opinions this is ok, or is it as morally complex as the stolen art debate.
BreakGPT:
…we’re all happily doomed…warm welcome to the final chapter in all sonic competition and consumerism…
ten years from now, we’re all enjoying listening to accoustic guitars while joining around campfires again…
at least those of us, who survived the big blackout…
never thought, i’d become a prepper one day…
let’s dance, one last time…
I think that sounds really good. I’m really not sure where I stand on the whole topic of AI in a creative space and I find it weird that anyone could be entirely pro or con - the various arguments have been reheased here and other threads many times so I’ll not delve into them…
But about that track, I agree that there are lots of really interesting textures and structural shifts and being honest, I’d happily listen to more of that sort of thing. And for me, it’s a much more interesting and creative song than 1000s of human songs because let’s face it… us humans are mostly just recycling and combining our own influences and true innovation is vanishingly rare. Not that innovation is essential it just make me think about all the times artists talk about how AI can’t replace their creativity.
Thanks for sharing it.
Same as AI art - not interested.