Is it time for SUPER plugin?

[Leo1999]

Oh I see,

[ztrauq]

It sounds like the concept you're discussing here is something similar to the shakeup that AI systems like Stable Diffusion are having in the world of visual art: an AI system that can hoover up a vast library of tagged music and create sonic "weights" that you could then use to apply a sonic character to a new piece of music that matches a certain musical style - the extreme of this would be somethign fully automated where you type in a text prompt and it generates a whole track on its own.

At that point, though, who is doing the composing - the AI? The musicians who created the tracks it was trained on and the styles it's now deriving/combining from them? I can type a prompt into DreamStudio and get a pretty picture back that I like, but it doesn't make me an artist.

Add to that, something in this vein would have the same problems of these types of AI algorithms: they can only act on the dataset they were trained on. They can create endless variations on the particular themes they've hoovered up, but they are completely incapable of creating anything truly novel. Doing that would require a quantum leap in our AI technology that at this point is not yet in evidence.

That being said, I think it is clear based on the direction of current VSTs that some level of AI tech will continue to see greater use in assisting musicians in their tasks, and can certainly have a profound impact in some of the more technical aspects of music - AI-assisted mastering, for example, is a sea change for the ability of small or independent artists to put out more professionally-tuned tracks. And even in the creative realm, there are any number of generators to help inspire or flesh out melodies or chord progressions.

Turning all of that into an interesting, coherent piece of music, though... aside from certain genres that are sufficiently rote that an algorithm may perhaps be all you need, I think AI is far away from being able to create something in the vein of this "super" plugin idea (and even that does seem to be described as imparting some creativity on the part of the producer tuning their initial music idea to a style, more in the vein of img2img than direct AI generation).

Never say never, though, and technology does keep getting better...

[whyterabbyt]

Oh I see

No, you dont.

[whyterabbyt]

It sounds like the concept you're discussing here is something similar to the shakeup that AI systems like Stable Diffusion are having in the world of visual art

It does, yes. But from the OPs posts its clear they've never actually tried these tools, nor do they understand them any deeper than what they've got from 'gee whiz' press articles, or have any historical perspective on their antecedents.

[Leo1999]

It sounds like the concept you're discussing here is something similar to the shakeup that AI systems like Stable Diffusion are having in the world of visual art: an AI system that can hoover up a vast library of tagged music and create sonic "weights" that you could then use to apply a sonic character to a new piece of music that matches a certain musical style - the extreme of this would be somethign fully automated where you type in a text prompt and it generates a whole track on its own.

At that point, though, who is doing the composing - the AI? The musicians who created the tracks it was trained on and the styles it's now deriving/combining from them? I can type a prompt into DreamStudio and get a pretty picture back that I like, but it doesn't make me an artist.

Add to that, something in this vein would have the same problems of these types of AI algorithms: they can only act on the dataset they were trained on. They can create endless variations on the particular themes they've hoovered up, but they are completely incapable of creating anything truly novel. Doing that would require a quantum leap in our AI technology that at this point is not yet in evidence.

That being said, I think it is clear based on the direction of current VSTs that some level of AI tech will continue to see greater use in assisting musicians in their tasks, and can certainly have a profound impact in some of the more technical aspects of music - AI-assisted mastering, for example, is a sea change for the ability of small or independent artists to put out more professionally-tuned tracks. And even in the creative realm, there are any number of generators to help inspire or flesh out melodies or chord progressions.

Turning all of that into an interesting, coherent piece of music, though... aside from certain genres that are sufficiently rote that an algorithm may perhaps be all you need, I think AI is far away from being able to create something in the vein of this "super" plugin idea (and even that does seem to be described as imparting some creativity on the part of the producer tuning their initial music idea to a style, more in the vein of img2img than direct AI generation).

Never say never, though, and technology does keep getting better...

I get what you are saying.

[jamcat]

you could even mix two or three artists,
imagine Paul McCartney and Sting for example

Would they be singing or playing bass?
Not playing drums, I hope!

[CinningBao]

I get what you are saying.

No, you aren't really hearing what we're saying.

[Leo1999]

I get what you are saying.

No, you aren't really hearing what we're saying.

Is wisdom science?

[kperry]

it's not a masturbation competition

How do you know who wins a masturbation competition?
Are points awarded for style?

Or is it a race?

Whoever comes first loses.

[whyterabbyt]

Its a shame, there are genuinely useful conversations to be had about AI processing, but this pie-in-the-sky take on it serving the laziest of all possible aspirations is about the least so.

[whyterabbyt]

Your wisdom is the only pure and right wisdom,
we all bow to your wisdom.
You simple don't get that it's just an idea

https://en.wikipedia.org/wiki/Psychological_projection

[Leo1999]

Its a shame, there are genuinely useful conversations to be had about AI processing, but this pie-in-the-sky take on it serving the laziest of all possible aspirations is about the least so.

And that's why you have joined???

[whyterabbyt]

Its a shame, there are genuinely useful conversations to be had about AI processing, but this pie-in-the-sky take on it serving the laziest of all possible aspirations is about the least so.

And that's why you have joined???

If you're setting the bar on your own efforts so low, why bother at all?

[Mind Riot]

How come it is accepted that drum loops are used, whether audio or midi files,
but if a plugin made a voice sound like Barry White or Michael Jackson
many a man would start complaining?!

If it was possible for a plugin to make me sound like Chris Cornell when I sang, I'd get it in a heartbeat and I'd tell anybody who had a problem with it what they could do with their opinions in florid rants of profanity. Does this make me a hypocrite? Yes, I think so. So I would happily go about my affairs as a hypocrite who could sing like freaking Chris Cornell, arguably the greatest rock singer ever in my opinion.

As a drummer of twenty five plus years, I can tell you that drum loops are NOT accepted by drummers. They were just brought into usage and they were easier to deal with than finding a competent drummer, so they proliferated. I feel the same way about them that guitarists will all feel when the tech finally reaches the point that they get replaced with MIDI just like so many drummers have been.

It's not all bad though. EDM isn't trying to sound like a real drummer and never has been, along with most electronically based music, so that stuff doesn't bother me; they're doing their own thing when it comes to percussion and natural sounding drums are not what they're after.

The straight up upside is that when people try to make MIDI grooves and loops into natural, authentic sounding drum performances most of them utterly fail. Just slapping down some MIDI is not going to glue your performances together; playing tight in a group and following the drummer are skills that you only get from playing with other people, and you can instantly tell when the musicians on a track either have that skill or don't.

It actually makes the skills of a real drummer more valuable, but most bedroom composers aren't going to have access to a solid drummer and they probably won't recognize the value one brings to the table.

Heck, I've even offered to play drums for free on some of my friend's projects and not gotten any takers, and I am a pretty damn good rock drummer who can play to a click comfortably and who has a pretty large catalog of work to demonstrate what I can do.

Oh well, lead a horse to water and all that, right?

[whyterabbyt]

I get what you are saying.

No, you aren't really hearing what we're saying.

I do wonder if the 'too busy to learn because I need to be on tik tok' aspirants will actually be happy spending hours a day listening to the results of their prompts, trying to get one that actually sounds decent.
At least Dall-E, Midjourney, SD etc produce results which can be assessed instantly, but not so a 30 second or 1 minute clip of music. Can you imagine them actually assessing a hundred variations on the same 'Paco De Luciafied' 30 second section?

Nah, because since Tik Tok, musicians dont have time to listen to their own music any more, right?

They'll need a SUPERSUPER plugin to do that bit for them, too.