I'm a relatively serious pianist and have been composing music (mostly classical/contrapuntal) since I was a kid.
A while back I wrote an article about how a musician could leverage GenAI along with the final piece.
The process I used was: start by singing to develop a motif → notate sheet music and chord progression → use an arranger keyboard to flesh it out → then convert the arrangement using the cover feature in Suno.
I think this is already possible with some manual effort.
However, AI companies likely avoid encouraging covers of copyrighted music to reduce legal risk. Record labels actively pursue unlicensed covers, and a company monetizing them would be an obvious target.
It would also likely make it easier to circumvent copyright detection algorithms.
I honestly wouldn’t call Suno “AI slop” at this point. Yes, there’s a ton of garbage coming out of it. But you can also create some really detailed and intricate songs if you do it right. I’ve made some really nice duets and trios for example. Cloned from a real voice, no… but I’m also not getting sued for stealing a real person’s voice.
And some of their upcoming features which aren’t generally known yet are going to be really useful to creators.
Slop is not about the quality of the rendering, it's about the lack of effort involved in generation. Generation is to creation as bubblegum is to food.
Surely art is about more than impressing you with my investment of effort, though? Otherwise you might as well go watch people working out at the gym as going to a musical production.
I think putting in low effort to make something which previously would have taken effort, is at high risk for producing slop. It tries to use the effort investment signal to suggest that you have something worthwhile to say, but the signal is fake.
But there are also works of low-key brilliance in the world long before AI, stuff that didn't take a lot of effort (and didn't pretend to) but were great anyway.
Slop is just the pejorative, and pejoratives always seize upon the biggest or most obvious weakness, which in the case of these models, IS in fact the quality of the results. But I agree that even if the quality somehow turns out great, the other major (and in my opinion more important) debate to be had is, at what point does technological assistance render the proximal human input irrelevant? And what is even the point of art if there was negligible human input?
Not a fair analogy. It’s all about the effort invested. If I hand AI “write me something funny about my dog” and get the song from that, slop. If I carefully detail my own lyrics and define line by line how I want it rendered, not slop. Unless my lyrics are terrible :)
Come on, do you think music critics are musicians by virtue of writing criticism? They write detailed descriptions of music with lots of context and emotional resonance, but I don't think that puts them in teh same category as their subject matter.
I'm a relatively serious pianist and have been composing music (mostly classical/contrapuntal) since I was a kid.
A while back I wrote an article about how a musician could leverage GenAI along with the final piece.
The process I used was: start by singing to develop a motif → notate sheet music and chord progression → use an arranger keyboard to flesh it out → then convert the arrangement using the cover feature in Suno.
https://mordenstar.com/blog/dutyfree-shop
I think this is already possible with some manual effort.
However, AI companies likely avoid encouraging covers of copyrighted music to reduce legal risk. Record labels actively pursue unlicensed covers, and a company monetizing them would be an obvious target.
It would also likely make it easier to circumvent copyright detection algorithms.
This cover has been going viral for some time:
https://www.youtube.com/watch?v=JSH0fXp4LoI
First ai-generated song that I had on repeat. I don't know the level of effort it required, but I don't care, honestly, it's very good.
I honestly wouldn’t call Suno “AI slop” at this point. Yes, there’s a ton of garbage coming out of it. But you can also create some really detailed and intricate songs if you do it right. I’ve made some really nice duets and trios for example. Cloned from a real voice, no… but I’m also not getting sued for stealing a real person’s voice.
And some of their upcoming features which aren’t generally known yet are going to be really useful to creators.
Slop is not about the quality of the rendering, it's about the lack of effort involved in generation. Generation is to creation as bubblegum is to food.
Surely art is about more than impressing you with my investment of effort, though? Otherwise you might as well go watch people working out at the gym as going to a musical production.
I think putting in low effort to make something which previously would have taken effort, is at high risk for producing slop. It tries to use the effort investment signal to suggest that you have something worthwhile to say, but the signal is fake.
But there are also works of low-key brilliance in the world long before AI, stuff that didn't take a lot of effort (and didn't pretend to) but were great anyway.
Slop is just the pejorative, and pejoratives always seize upon the biggest or most obvious weakness, which in the case of these models, IS in fact the quality of the results. But I agree that even if the quality somehow turns out great, the other major (and in my opinion more important) debate to be had is, at what point does technological assistance render the proximal human input irrelevant? And what is even the point of art if there was negligible human input?
Not a fair analogy. It’s all about the effort invested. If I hand AI “write me something funny about my dog” and get the song from that, slop. If I carefully detail my own lyrics and define line by line how I want it rendered, not slop. Unless my lyrics are terrible :)
It’s a tool. The question is how you use it.
Come on, do you think music critics are musicians by virtue of writing criticism? They write detailed descriptions of music with lots of context and emotional resonance, but I don't think that puts them in teh same category as their subject matter.