What's new

SPINOFF: Made with AI: Suno, Udio and others - Discussion

PedroPH

Active Member
MODERATOR NOTE: This thread was spun off from a thread originally designed to share the results of using AI generation tools. That thread can be found here: https://vi-control.net/community/threads/made-with-ai-suno-udio-and-others.150986/post-5527670

==================================

Why is it sad? Isn't it great that music can be composed with little effort? This will be hard on some people for a while, but in time people will adapt.

Also, we shouldn't think in black and white. It's not "music painstakingly composed by a human by playing or entering each note" vs "music generated automatically with a prompt and the push of a button". I think soon enough we'll see great tools for humans to compose with a lot less work. It will take more time than simply pushing a button, but the music will be completely composed by a human, only with a lot less effort. Imagine if you could hum a melody and ask the software to play it as if a choir were singing it. Then you hum the accompaniment. You don't need to edit CC automation curves anymore, pick a sample library, etc. You just hum the melody and say "I want this melody played by first violins at 120 BPM". The software creates it. "I want it a little faster". "I want it in a larger hall". "Now add second violins playing one octave below". Etc.

Perhaps music software producers should be more worried than composers.

EDIT: Re-reading this, I realize it would be better "isn't great that music can be PRODUCED with little effort". What I means is gains in efficiency and productivity are good. To do more with the same or fewer resources is good. But then, as my argument progresses, "composed" becomes more suitable.
 
Last edited by a moderator:
Why is it sad? Isn't it great that music can be composed with little effort?
No.
Why is it great? No one is composing anything.
Just writing prompts and pressing enter.
No creativity skill or talent needed.

This will be hard on some people for a while, but in time people will adapt.

Also, we shouldn't think in black and white. It's not "music painstakingly composed by a human by playing or entering each note" vs "music generated automatically with a prompt and the push of a button". I think soon enough we'll see great tools for humans to compose with a lot less work. It will take more time than simply pushing a button, but the music will be completely composed by a human, only with a lot less effort. Imagine if you could hum a melody and ask the software to play it as if a choir were singing it. Then you hum the accompaniment. You don't need to edit CC automation curves anymore, pick a sample library, etc. You just hum the melody and say "I want this melody played by first violins at 120 BPM". The software creates it. "I want it a little faster". "I want it in a larger hall". "Now add second violins playing one octave below". Etc.

Perhaps music software producers should be more worried than composers.
Yes, it could be nice to hum a melody for the first violins then the second violins, then hum a cello line etc etc and the thingy makes it all sound like the real deal and making us skip choosing the library, programming velocity curves, fixing timing, mixing etc...

The fact is that for now this is not the aim of the tech.

The aim is just to cut you musician completely off. F..,,,k you up, make you disappear, make you redundant...
it is not made to be a tool for musicians, it is a tool for anyone so they don't need to use a musician.
 
Why is it sad? Isn't it great that music can be composed with little effort?
“Composed” is poorly chosen here. Maybe “created.” In any case insofar as one is working with AI feeding it prompts the activity is closer to being a music editor working with a set of libraries trying to find the right track and adapting it. Or if you are making deeper interventions it’s akin to arranging or producing a track of existing music and audio. There can be an element of composition to this. But it’s not the primary labor input.

And yet there are in fact so many aspects of composition that are at base algorithmic and so can just as well be done by the machine. Look at how so many folks use drum loops and automated drummers. Or stock accompaniment patterns, simple ostinatos, arpeggiator presets.Then too many use commercial loops creatively and I see no reason that they couldn’t use AI produced loops equally effectively. The line gets very fuzzy indeed.
 
“Composed” is poorly chosen here. Maybe “created.” In any case insofar as one is working with AI feeding it prompts the activity is closer to being a music editor working with a set of libraries trying to find the right track and adapting it. Or if you are making deeper interventions it’s akin to arranging or producing a track of existing music and audio. There can be an element of composition to this. But it’s not the primary labor input.

And yet there are in fact so many aspects of composition that are at base algorithmic and so can just as well be done by the machine. Look at how so many folks use drum loops and automated drummers. Or stock accompaniment patterns, simple ostinatos, arpeggiator presets.Then too many use commercial loops creatively and I see no reason that they couldn’t use AI produced loops equally effectively. The line gets very fuzzy indeed.
I was referring to the generation of the music itself, i.e. what the machine does, not what the person writing the prompt does. A person that writes a prompt so that the machine generates music isn't composing music at all, the same way a film director asking a composer to write some cheerful music isn't composing. But what a machine does after receiving a prompt in order to generate music, that I think it's permissible to call "composing". "Creating" works too.
 
I was referring to the generation of the music itself, i.e. what the machine does, not what the person writing the prompt does
Sure, if you want to play at being a director or music editor, it can maybe be a fun game.

The problem with your initial formulation was that it potentially confuses writing prompts with writing music. I also don’t know what I’d call the activity that the AI engages in. I guess I’d say it is creating material. Composition also involves creating materials. But I still don’t know if I’d call AI work composition any more than I’d call any other bit of algorithmically generated music composition. Algorithms generate material but can an algorithm generate a composition in whole? I suppose that’s a question to be put to total serialist and fully aleatoric pieces as well. (Once upon a time there was quite a lot of discussion about such things, so they are not new questions.)
 
Impressive leap in audio over Suno, but I wouldn't jump to conclusions before working more closely with it, since Suno has proved itself with some good understanding of each style, and how melody reflects the text, Udio has some big boots to fill.

Is it ethical? I've yet to see evidence they used unlicensed materials, and it seems like lots of retro/classical stuff was used, so I assume its licensed from smaller recording labels?

I was hesitant to "take" Suno as collaborator, but it proved to be very nice and interesting experience - specially since its so attentive to the lyrics you insert, and the way you format it is very important. I felt legitimate lyricist for a writing companion, and not a monkey trying to pull one handed vegas automate and see what happens. I am yet to try Udio, so looking forward to use it once time allows...
 
Last edited:
I just generated a tragic folk song about eating a marshmallow and it's better than anything I could've hoped to make myself.

I guess it's a good thing I picked teaching as a career instead of composer...who am I kidding, AI will take that job too.
 
I picked teaching as a career instead of composer...who am I kidding, AI will take that job too.
And the students’ “job” as well. Right now it seems a race between AI paper writing and assignment completion and AI grading, which university administers are very keen on implementing since grading is very labor intensive and the main thing keeping instruction from scaling. The logic to the situation is easy enough to game out: AI completed assignments graded by AI “instructors.” Which should give everyone involved plenty of time to listen to their allocation of AI generated music and generate daily episodes of AI produced TV shows and movies. Probably even AI generated cute cat videos. Should be a fun time!
 
Is it ethical? I've yet to see evidence they used unlicensed materials, and it seems like lots of retro/classical stuff was used, so I assume its licensed from smaller recording labels?
It's pretty clearly using unlicensed material. A recording label of classical music and opera isn't going to allow their catalogue to be used for AI, if only for the legal headache of getting all the performance rights signed over. That world is also the least likely to adopt AI.

There's nowhere near enough production library opera to get this quality. Just orchestral would be similarly difficult- all public domain recordings of this style are pretty bad quality, unsuitable for training. It's also not a popular style found in most production libraries, so you'd need to incorporate a big number of libraries to get enough data.

Or, y'know, just rip it from Spotify and off you go!
 
Sure, if you want to play at being a director or music editor, it can maybe be a fun game.

The problem with your initial formulation was that it potentially confuses writing prompts with writing music. I also don’t know what I’d call the activity that the AI engages in. I guess I’d say it is creating material. Composition also involves creating materials. But I still don’t know if I’d call AI work composition any more than I’d call any other bit of algorithmically generated music composition. Algorithms generate material but can an algorithm generate a composition in whole? I suppose that’s a question to be put to total serialist and fully aleatoric pieces as well. (Once upon a time there was quite a lot of discussion about such things, so they are not new questions.)
For now algorithms can only generate small chunks of music, I think. But more complex things are a matter of time, I guess.
 
No.
Why is it great? No one is composing anything.
Just writing prompts and pressing enter.
No creativity skill or talent needed.
True. But it's still music. And it's produced at a fraction of the cost. Perhaps it's not as good. But it has its uses and can improve the lives of many people (e.g. people who need music for their videos and can't afford to hire a composer).
No.
Why is it great? No one is composing anything.
Just writing prompts and pressing enter.
No creativity skill or talent needed.


Yes, it could be nice to hum a melody for the first violins then the second violins, then hum a cello line etc etc and the thingy makes it all sound like the real deal and making us skip choosing the library, programming velocity curves, fixing timing, mixing etc...

The fact is that for now this is not the aim of the tech.

The aim is just to cut you musician completely off. F..,,,k you up, make you disappear, make you redundant...
it is not made to be a tool for musicians, it is a tool for anyone so they don't need to use a musician.
No. The aim is to be able to produce music more easily. Or to lower costs. The aim of mass production of shoes wasn't to damage shoemakers. Was it?
 
For now algorithms can only generate small chunks of music, I think. But more complex things are a matter of time, I guess.
It’s not a question of complexity. These problems of algorithmic production of music have long been discussed and not really resolved. I think no one has issues with algorithm as generating material that is then worked on by the composer. But the closer the situation gets to the algorithm generating the piece the more doubt there is as to whether the generated output is a “composition.” You might take from that that there is a definitional sense that composition as such requires the input of human labor. Which of course might ultimately be a distinction that is completely moot, or invested in a nostalgic hope that humanity has not rendered itself superfluous to its machines.
 
True. But it's still music. And it's produced at a fraction of the cost.
KFC is still chicken
Perhaps it's not as good.
Might be actually better who knows...
But it has its uses and can improve the lives of many people (e.g. people who need music for their videos and can't afford to hire a composer).
You pay a few $ a month and have so many commercial options already for this use, microlicensing etc....
No. The aim is to be able to produce music more easily.

Why produce music if you are not actually producing?

AI art is NOT an act of creation of the human "prompter".

Hey I encountered people that were content to write the song title and a couple of lyrics, let everyone else do all the writing, playing, recording production etc, pay a nice buyout and then claim the music as theirs.

So I guess not every one is interested in the act of making art and are content with claiming a vague and weird ownership.
 
In the first Industrial Revolution they came for our crafts in the second they are coming for our arts.

Anyone who believes these algorithms are being created to serve the artists needs to study their history.

As Cloud technology lead for a large university my job is safe for the foreseeable but never in the 30 years of my career have I wanted my job less. One day we may see AI produce the cure for cancer or solve our ever growing energy needs. Yet daily the work/research is see being done for the benefit of humanity feels like such a small slice of the whole.
AI and those working in the field have a moral question to ask themselves every day. Sadly there are too many immoral people in the world who see success in $€&£ value at all costs as their goal.

Someone above said what a time to be alive, certainly it is, that doesn’t mean it’s a good time. Only we can steer the direction we take our future in and farming out our ability to create is not the direction I ever want to head in.
 
It’s not a question of complexity. These problems of algorithmic production of music have long been discussed and not really resolved. I think no one has issues with algorithm as generating material that is then worked on by the composer. But the closer the situation gets to the algorithm generating the piece the more doubt there is as to whether the generated output is a “composition.” You might take from that that there is a definitional sense that composition as such requires the input of human labor. Which of course might ultimately be a distinction that is completely moot, or invested in a nostalgic hope that humanity has not rendered itself superfluous to its machines.
Yes. That is just a matter of semantics. Besides, composition required human labor because machines couldn't do it. But now that they can generate music... It's debatable, though. I don't think it right to say that a machine learning software "understands" text. Even the "learning" in machine learning seems a misnomer. It doesn't learn. It just calculates weights... But then, as a metaphor, "learning" is ok...
 
Yes. That is just a matter of semantics. Besides, composition required human labor because machines couldn't do it. But now that they can generate music... It's debatable, though. I don't think it right to say that a machine learning software "understands" text. Even the "learning" in machine learning seems a misnomer. It doesn't learn. It just calculates weights... But then, as a metaphor, "learning" is ok...
Yes, LLM AI recognizes patterns. It generates new items through random variation of reproducing patterns. There's an affinity with a certain mode of human creativity in this, which is why I don't think it can be completely dismissed and it might indeed turn out to be very good at making materials, simply because quantity often yields quality, and the one thing it is already very good at is delivering in quantity...

I would say composition required human labor because that conceptual human labor was deemed integral to its standing as music. This is why there have been debates about algorithmic "composition" and whether it was even music since it came onto the scene. Those debates were never definitively decided, and I imagine the reason for that has to do with ambiguities that reside in the ontology of music. I don't really expect AI to change that, though I expect it will bring a great deal of legal clarity to the issue of algorithms and copyright.
 
Yes, LLM AI recognizes patterns. It generates new items through random variation of reproducing patterns. There's an affinity with a certain mode of human creativity in this, which is why I don't think it can be completely dismissed and it might indeed turn out to be very good at making materials, simply because quantity often yields quality, and the one thing it is already very good at is delivering in quantity....

But Jim, that “affinity with human creativity“ is not just an affinity, but a modular cross-referencing of existing products of human creativity. Its direct causal precedent is the creativity located in its input training set and nowhere else. And it is solely the quantity and quality of that input which determines the quality of any amount of output.

What needs to be borne in mind – and granted, it is difficult to wrap one’s head around – is the sheer breathtaking scale of the theft transpiring here.

What people might refer to as analogy is e.g. the purchase of Lucasfilm’s material in Star Wars by the Disney corporation. Disney reserves the right to, by whatever means, iterate over and create new variations on existing works. The deal was eyewateringly expensive. So should these ventures be. That is a matter of organization and channeling of power, whatever power music makers may still have en masse.
 
Last edited:
The continuous appraisals in multiple threads - from musicians who should now better - of this utter, utter, piece of shit software is simply astounding. I shake my head in disbelief. You’re digging your own musical graves, for f*** sake !!
How is this a bad software exactly? And besides, what comes first, good music (which current models occasionally achieve) or "Me!"....
 
True. But it's still music. And it's produced at a fraction of the cost. Perhaps it's not as good. But it has its uses and can improve the lives of many people (e.g. people who need music for their videos and can't afford to hire a composer).

No. The aim is to be able to produce music more easily. Or to lower costs. The aim of mass production of shoes wasn't to damage shoemakers. Was it?
If you would be talking about gardening tools I would totally agree. And by the way, it probably wasn‘t the aim to damage shoemakers, but it was the result !
 
Top Bottom