I think the computers just won.

Yeah it’s going to be amazing in some months. 

What is needed from Ai is a much quicker way to get your ideas down, and to realize them FULLY. Would love to be able to give it a short one shot, then with a prompt as it to generate tons of variations. I mean I already kinda do that when you throw FX on existing sounds, record it, resample it, etc. I guess it would just be quicker. If I have a specific idea in mind then any tool can only bring that idea about quicker.

We also need private Ai, in that we are guaranteed what we might enter into the Ai to alter isn't then available to everyone else. So we end up with our own Ai generated audio libraries that WE own. Say we put a ton of samples we made in some DAW, into the Ai and ask it to mix and match and generate similar samples. We need a guarantee that none of it is available for re-use by others. Where as soon as your synchronize your audio, it's available for everyone else to use. The end product might since it can be copied, but all the materials That went into it's creation, all the iterations, generated samples, prompts, etc  etc etc, remain ours privately to use and re-use later as we see fit.

Currently however it craps out watered down replicas of probably stolen copyrighted material and free audio materials. For any ai generating software, there should be a master list of all materials that go into it, and if a person chooses so they can have their music taken out of the training materials ( which is a euphemism tech bros use for stolen audio materials). Tho it shouldn't be there in the first place. 

Ai can be a great tool IF the artists are protected. It needs to work for us, not replace us. But when you live in a corporate hellscape where money is the bottom line, I have little hope for any of it. Ai is currently on track to dilute humanities potential when it could instead bolster it. It's just that people have to give a shit for the second one to come true.

  On 4/15/2024 at 9:32 PM, Brisbot said:

replicas of probably stolen copyrighted material and free audio materials

Absolutely, this is the thing that 'AI artists' don't quite seem to have grasped - Essentially it's just a glorified version of a search engine that just mulches the search results.

Whenever there's a big enough copyright case brought forward all this stuff is gonna go back into hiding. We're basically at the free-for-all equivalent of when Napster first came out - Maybe Metallica will be the people that try to get all shut this down too

  On 4/15/2024 at 9:50 PM, mcbpete said:

Absolutely, this is the thing that 'AI artists' don't quite seem to have grasped - Essentially it's just a glorified version of a search engine that just mulches the search results.

Whenever there's a big enough copyright case brought forward all this stuff is gonna go back into hiding. We're basically at the free-for-all equivalent of when Napster first came out - Maybe Metallica will be the people that try to get all shut this down too

Expand  

Google does the same with their models for serving ads. They delete the trained data on your person but the model stays trained 

  On 4/15/2024 at 8:24 PM, hma said:

Everything I`ve heard so far has this terrible spectral smearing all over it and I don`t think it`s the encoding.

i assumed that was on there because it’s being trained on tons of low quality rips/feeds…everything’s got that washy mp3 sound to it, super noticeable in the high hats and stuff.

that'll be gone within a few iterations/new companies upping the quality of data tho.

  On 4/15/2024 at 4:33 PM, mcbpete said:

I tried a prompt injection to see if I could get it to make a track with lyrics citing where it's getting its music and lyric training data from.... completely unsuccessful but without any other prompting I got this:

https://www.udio.com/songs/nZjnnLFxyHi5utwb3qVGWY

holymaloney ... After the first 45 seconds of late 90s cyber-hacker-core it hits HARRDD, so just hit continue until it hit the track length limit - I'd legitimately want to hear albums of this!

Now I'm even more curious as to where the training data has come from.

Expand  

Yeah this is great. After making loads of dumb stuff I sat at the weekend trying to make something interesting and I was really surprised with what it could come up with. I started off trying to see if it could make some 2008 era dubstep stuff, but ended up picking different bits until it made this weird wobbly, sad sounding thing that I legit enjoy:

https://www.udio.com/songs/nKEuScu2ecceVPNqrmj4vr

It's a cracking bit of technology. Really excited to see where it goes.

Come on you cunts, give me some slow Zelah amen break, ok?

gettin some good drones 

https://www.udio.com/songs/qXfWW4jAsqBJs3HfBVYfrx

https://www.udio.com/songs/myU3vNkYoXSctDNUhZrAt9

  On 4/15/2024 at 8:55 PM, mcbpete said:

I'm presuming its just a generational thing from [fft block?] size of the training data, remember when AI image generators only just over 5 years back were only capable of smushy abstract jpgs

Expand  

This looks pretty ominous lol

  On 4/12/2024 at 3:30 PM, chronical said:

holy shit lol

https://www.udio.com/songs/s7m6zMX9T7U3Xes1Q8C1hH

Lol so that's how rhythm and sound does it

https://www.udio.com/songs/sNEXWUoWGydjU2RkhnC4rv

gangsta rapper has a small package 

Peter Murphy passing a kidney stone 

https://www.udio.com/songs/37YD7ZicwCzCoL7kEew6Ze

John F. Kennedy "raps" about the Grassy knoll

jazzy collage into amenbreakfukery (prompt was for the outro lol)

acid tune to ab4 techno

hiphop beat

gabor lazar vs ae

Wings meets Alan Partridge

https://www.udio.com/songs/bi2tfEquztNtE4yvKcw9Ep

experimented with soundtracking some of my old dalle2 artwork with Udio, as a sci-fi pop song.  I'm fascinated by art using Ai for both the sound and the visuals

https://www.udio.com/songs/shCHEnbvY1dxBdyZDqne8m

  On 4/26/2024 at 2:48 AM, Rubin Farr said:

experimented with soundtracking some of my old dalle2 artwork with Udio, as a sci-fi pop song.  I'm fascinated by art using Ai for both the sound and the visuals

https://www.udio.com/songs/shCHEnbvY1dxBdyZDqne8m

Expand  

Pretty nice stuff my friend. With this I do see how creativity can be fostered with these tools.

  On 4/26/2024 at 3:55 AM, logakght said:

Pretty nice stuff my friend. With this I do see how creativity can be fostered with these tools.

It really is only limited by your imagination, and this is just the initial version of this type of model. Very exciting time imo.

https://www.udio.com/songs/mXcgWJfF2HXWKEgtTnJzAk

Subway bathroom devil's 3-way ends in baby daddy drama

https://www.udio.com/songs/17ogZ1uPMi6S1XybDYH34U

"love's graffiti, sprayed then wiped away."  

  On 4/16/2024 at 10:23 AM, Blir said:

Yeah this is great. After making loads of dumb stuff I sat at the weekend trying to make something interesting and I was really surprised with what it could come up with. I started off trying to see if it could make some 2008 era dubstep stuff, but ended up picking different bits until it made this weird wobbly, sad sounding thing that I legit enjoy:

 

https://www.udio.com/songs/nKEuScu2ecceVPNqrmj4vr

 

It's a cracking bit of technology. Really excited to see where it goes.

Expand  

this sounds like current-era Amon Tobin.

vodka vodka boris yeltsin (demo for трип)

good thoughts here

but i'm honestly not sure if he's quite right. if someone trains some AI/ML prog with enough jam sessions (like, man, idk, the tens of thousands of hours of recordings that already exist of live jazz, Grateful Dead/Phish/etc., folk, etc....) then you could probably get some AI jamming pretty well.

there's more to his argument tho. music as a verb, music as a collective human experience & conversation, players being constrained and influenced by their physical bodies, and so on. i thought it was a really nice take.

  On 5/1/2024 at 1:28 AM, auxien said:

good thoughts here

but i'm honestly not sure if he's quite right. if someone trains some AI/ML prog with enough jam sessions (like, man, idk, the tens of thousands of hours of recordings that already exist of live jazz, Grateful Dead/Phish/etc., folk, etc....) then you could probably get some AI jamming pretty well.

there's more to his argument tho. music as a verb, music as a collective human experience & conversation, players being constrained and influenced by their physical bodies, and so on. i thought it was a really nice take.

Expand  

What AI will never understand is the "symbolic" and "imaginary" dimensions of art creation. It is as real as you and me creating music, but it lives in another realm of "enjoyment".

I'm impatient and didn't watch that video but did Neely mention George Lewis' Voyager stuff in this context because it seems relevant

But yeah, I dunno about the genuine musical potentials of any of the AI stuff but I do like that we're in a moment right now where I can get it to make weird noises like this https://www.udio.com/songs/myHPz4ZD5tJE5VZiZG8BNb

A user posted this on Reddit 

https://chat.openai.com/g/g-UVazgQMhI-udio-music-creator-assistant

Beltane Greetings

https://www.udio.com/songs/bZaUqtavEp4vvHJWueT1Nq