Menu
What's new
By:
Filters

UDIO AI- short serial piece for baritone and orch

I'm guessing that not long from now there'll be a whole lotta people passing off marginally edited AI excretions as their own music. And when it becomes rather well known that some people are doing that, then composers will get to enjoy the thrill of having their own authentic music being held in suspicion as possibly being the result of AI. You'll have to have video of the entire process of composition just to have half a chance of proving that what they're hearing is your baby.

I'm also guessing that at some point AI will be officially declared its own race among the human races and you won't be able to say anything against it or else you're an anti-AI racist and the AI censorship algorithms will erase your entire digital existence.

Clown World will not disappoint, cuz it never does.
__________
 
tc9000 said:
The prompt was the first line: modern classical, serialism, atonal, contemporary classical, haunting ending

I think the second line is some sort of categorisation? Not sure
Click to expand...
Where do you see the text... "modern classical, serialism, atonal, contemporary classical, haunting ending"

I don't see it.
 
StefanoM said:
Just another example...of copyright problem on this AI technology...

THIS IS of course a TRAINING on ANDREA BOCELLI !
And this is NOT ACCEPTABLE!"

There is nothing exciting or beautiful in all this. Considering that this is NOT making music.
Click to expand...
Agreed - this is copyright-infringement on a massive scale. But will the courts side with the copyright owners, or will there be tremendous pressure from very wealthy AI corporations/funders to let this slide (for the sake of national security, and other excuses)?

We will see. My guess - the corporations will win, copyright legislation will be revised, and copyright owners will get very meager compensation.
 
soundofmaw said:
Agreed - this is copyright-infringement on a massive scale. But will the courts side with the copyright owners, or will there be tremendous pressure from very wealthy AI corporations/funders to let this slide (for the sake of national security, and other excuses)?

We will see. My guess - the corporations will win, copyright legislation will be revised, and copyright owners will get very meager compensation.
Click to expand...
The owners of the copyrights in this case are also for the most part very wealthy corporations as well, so I don't think it's so easy to predict which of our very large business concerns the courts are likely to side with on those grounds. Precedent is on the side of the copyright holders, though, so if I was going to place a bet that's where I'd place it (and Google lost to the authors on its attempt to digitize all the books).
 
soundofmaw said:
Agreed - this is copyright-infringement on a massive scale. But will the courts side with the copyright owners, or will there be tremendous pressure from very wealthy AI corporations/funders to let this slide (for the sake of national security, and other excuses)?

We will see. My guess - the corporations will win, copyright legislation will be revised, and copyright owners will get very meager compensation.
Click to expand...

I hope you’re wrong, but I think you’re right.

It’d be such sweet irony to train an AI model based on Suno and Udio’s models, sell it for cheaper to eat their market share, and then tell them that you’re "democratizing" their industry when they cry about it.
 
one must not underestimating other industries.

It's not like these are GOD. If the entire world of art, culture, entertainment, and music, big lawyers, moves against all this, something will happen.

Don't forget that LOTS of money also circulates elsewhere. And there are just as many powerful people who govern the world of music. Then there is an ethical issue at the core that is important.

This stuff, as it is... benefits no one. Except for those who make money from subscriptions.

The music you make with these means, are public domain, they are not yours, the output can be similar to another's.

They are profoundly useless tools, but they risk making a lot of money, indeed, by STEALING talent, information, art, without anyone's permission. And it's not just important names that have ended up inside, but also less KNOWN people, unaware of having given their talent to them to create a data set.
 
jbuhler said:
The owners of the copyrights in this case are also for the most part very wealthy corporations as well, so I don't think it's so easy to predict which of our very large business concerns the courts are likely to side with on those grounds. Precedent is on the side of the copyright holders, though, so if I was going to place a bet that's where I'd place it (and Google lost to the authors on its attempt to digitize all the books).
Click to expand...
Good point. I think the difference between this and what Google was doing with books is this is a bit of an "arms race" so to speak against other nations (ie. China, Iran, Russia, etc.) that are also rapidly developing AI. Western governments (and courts) may take that into consideration when ruling on these cases. On the other hand, this affects a lot of wealthy and no-so-wealthy artists/creators that those same governments are accountable to. I think a compromise will be reached eventually. I just hope the compensation for REAL creators won't be the pittance that I think it might be.
 
This is all straight out of the Google playbook - do the thing and let the lawyers fight about it afterwards. They are moving so fast, the law is years behind the technology 😂.

I'm so conflicted on this. As always, AI offers shiny, fascinating, terrrifying toys to play with, but these may come with a cost, and we don't know what that will be.

I hope that, if nothing else, the current virtual vocalist technology can be improved upon. I'm less interested in AI making music and more interested in making music with AI-enhanced tools. If someone could implement the pitch, note, timbre, dynamics etc control that we have in VSTs and just harness this quality of these artificail voices, that would be awesome! But I fear that either the tech wont be capable of being untangled in that way, or more likely noone will care about this albeit incredibly niche area of wanting improved vocal VSTs, and people will care more about being able to 'produce' music where you just write a prompt.
 
Last edited:
N.Caffrey said:
this is 100% Pavarotti!! wtf?!?
Click to expand...
Agreed - I didn't even realise when I generated that - I just thought - that sounds shockingly real and far better than any of the virtual vocalists we currently have... but yeah - fully agreed. Reminds me a bit of that macabre thing where Star Wars used a digital likeness of Peter Cushing after he passed away
😬
 
So here is an idea, probably too naive mind you: feed in say every recording from the London Symphony Orchestra. Then make it only available as a one time purchase for rendering from a notation program or DAW. Or offer a subscription which the revenue goes towards the institution that AI “sampled”.

I’ve mentioned this numerous times but the only value I see in AI as a composer is improved playback of a score. It won’t replace the sample libraries or NotePerformer for the process of creation but only the final render when you either have to deliver a score to a conductor or I suppose a client (when the budget isn’t there for real musicians and hey let’s be honest, this latter option has been happening since the ‘80s with technology).
 
dcoscina said:
So here is an idea, probably too naive mind you: feed in say every recording from the London Symphony Orchestra. Then make it only available as a one time purchase for rendering from a notation program or DAW. Or offer a subscription which the revenue goes towards the institution that AI “sampled”.

I’ve mentioned this numerous times but the only value I see in AI as a composer is improved playback of a score. It won’t replace the sample libraries or NotePerformer for the process of creation but only the final render when you either have to deliver a score to a conductor or I suppose a client (when the budget isn’t there for real musicians and hey let’s be honest, this latter option has been happening since the ‘80s with technology).
Click to expand...
Yes, I agree.

These are two things that can and should continue to exist.

The creative process must not be replaced. I continue to believe that only those who do not love making music can find tools like UDIO useful, as they are totally destructive to the concept of music itself.

It's unlikely that AI systems, at least with today's computers, will be able to operate in real time.

So yes, I find this reasoning useful.

Classic VI and Sample Libraries are necessary tools for the creative, compositional, sound design process, also because I like to PLAY and music, when I compose I like to play the parts, not just have them in my head. While in the future I will find useful an AI capable of helping me with the final rendering.

And maybe structure the scores quickly, perhaps starting from the analysis of my mockup. But the composition would have been done by me, I would have composed, playing the parts, enjoying doing so. Just as it happens right now.
 
StefanoM said:
Yes, I agree.

These are two things that can and should continue to exist.

The creative process must not be replaced. I continue to believe that only those who do not love making music can find tools like UDIO useful, as they are totally destructive to the concept of music itself.

It's unlikely that AI systems, at least with today's computers, will be able to operate in real time.

So yes, I find this reasoning useful.

Classic VI and Sample Libraries are necessary tools for the creative, compositional, sound design process, also because I like to PLAY and music, when I compose I like to play the parts, not just have them in my head. While in the future I will find useful an AI capable of helping me with the final rendering.

And maybe structure the scores quickly, perhaps starting from the analysis of my mockup. But the composition would have been done by me, I would have composed, playing the parts, enjoying doing so. Just as it happens right now.
Click to expand...
Totally agree. 100%. The fun of composing is discovery. It helps us grow as composers. If that element is removed then we really cannot call ourselves composers otherwise. More like arrangers or less..
 
dcoscina said:
Totally agree. 100%. The fun of composing is discovery. It helps us grow as composers. If that element is removed then we really cannot call ourselves composers otherwise. More like arrangers or less..
Click to expand...
The process feels most like editing music to picture, that is, music editing. But without the picture. You do have to have an idea where you want the music to go as you prompt it. But even then the bits it spits back often don’t quite go together so you have to edit….

I played with Udio some yesterday and i was completely unimpressed with Udio’s response to prompts. For orchestral music, at best it gets large genre prompts right but it rarely gets even close with details such as tempo, mood, or instrumentation. It would be hilarious seeing a director or music editor trying to use this to score to picture. Maybe Udio needs much better tagging data.

The sound is the most impressive thing about Udio but so much of its output sounds like it has been cobbled together from other recordings, and the case law on unauthorized sampling at this point is deep. It’s a copyright infringement suit waiting to happen if it’s ever used commercially at this point. I really don’t expect to see the recording companies and music publishers to lose that, meaning the tech bros will need to buy out the recording and publishing companies, pay to license the content, or manage to lobby to get laws changed. While they might make a national security case for text, voice, image, and video, because of how news and information is disseminated through them, it’s much harder to see how they make that case for music.

And if they want to say that people are also trained on existing music, people are also held to a high standard of not infringing with their knowledge, and composers and authors generally have to indemnify publishers snd clients and certify that the work does not infringe when they sign contracts. I can’t imagine large media concerns would treat music from AI differently in terms of risking lawsuits, and if the law is changed to allow what would now be infringement, humans will have to be allowed to similarly infringe.
 
jbuhler said:
The process feels most like editing music to picture, that is, music editing. But without the picture. You do have to have an idea where you want the music to go as you prompt it. But even then the bits it spits back often don’t quite go together so you have to edit….

I played with Udio some yesterday and i was completely unimpressed with Udio’s response to prompts. For orchestral music, at best it gets large genre prompts right but it rarely gets even close with details such as tempo, mood, or instrumentation. It would be hilarious seeing a director or music editor trying to use this to score to picture. Maybe Udio needs much better tagging data.

The sound is the most impressive thing about Udio but so much of its output sounds like it has been cobbled together from other recordings, and the case law on unauthorized sampling at this point is deep. It’s a copyright infringement suit waiting to happen if it’s ever used commercially at this point. I really don’t expect to see the recording companies and music publishers to lose that, meaning the tech bros will need to buy out the recording and publishing companies, pay to license the content, or manage to lobby to get laws changed. While they might make a national security case for text, voice, image, and video, because of how news and information is disseminated through them, it’s much harder to see how they make that case for music.

And if they want to say that people are also trained on existing music, people are also held to a high standard of not infringing with their knowledge, and composers and authors generally have to indemnify publishers snd clients and certify that the work does not infringe when they sign contracts. I can’t imagine large media concerns would treat music from AI differently in terms of risking lawsuits, and if the law is changed to allow what would now be infringement, humans will have to be allowed to similarly infringe.
Click to expand...
I agree.

It's evident that currently, these tools are only viable for pure curiosity. Achieving exactly what I want seems unattainable.

For this reason, it is challenging for me to envision using them in a professional setting, where accurately interpreting client requests is already demanding.

Moreover, you must ensure that the AI comprehends the demands of the client, the director, and the supervisor, communicated through a prompt.

Suppose you're fortunate enough to find the right combination. What happens if there's a request for a modification? You're out of luck.

Fundamentally, these tools are designed to wow audiences, but as they stand, they are largely ineffective for anything beyond creating random, unfocused outputs.
 
jbuhler said:
The process feels most like editing music to picture, that is, music editing. But without the picture. You do have to have an idea where you want the music to go as you prompt it. But even then the bits it spits back often don’t quite go together so you have to edit….

I played with Udio some yesterday and i was completely unimpressed with Udio’s response to prompts. For orchestral music, at best it gets large genre prompts right but it rarely gets even close with details such as tempo, mood, or instrumentation. It would be hilarious seeing a director or music editor trying to use this to score to picture. Maybe Udio needs much better tagging data.
Click to expand...
I noticed that too. Asked for only drums and vocals and a whole band was playing etc etc...
But.

It's been out for literally days....

UNFORTUNATELY give it time and it will respond better to prompts. It will adjust to sync points, and the director's (prompter's) instructions about where to go intense or where to go more unnoticed, what to convey on this or that point etc etc....

jbuhler said:
The sound is the most impressive thing about Udio but so much of its output sounds like it has been cobbled together from other recordings, and the case law on unauthorized sampling at this point is deep. It’s a copyright infringement suit waiting to happen if it’s ever used commercially at this point.
Click to expand...
Yes I agree, but how does it put it together so well?

I clearly heard Pavarotti, Bocelli, Brel, Nusrat Fateh Ali Khan, Victor Jara and many others....
(BTW what is frighting is also that sometimes they give me the impression to really "mean" the lyrics sang, I swear I ain't going crazy yet. There is an appearance of a connection between lyrics and the performance sometimes I found.)

They are just putting it out, infringing so many rights in the process, without any care as they plan to pass all law issues to the user's end.
It's scandalous.
 
dcoscina said:
So here is an idea, probably too naive mind you: feed in say every recording from the London Symphony Orchestra. Then make it only available as a one time purchase for rendering from a notation program or DAW. Or offer a subscription which the revenue goes towards the institution that AI “sampled”.

I’ve mentioned this numerous times but the only value I see in AI as a composer is improved playback of a score. It won’t replace the sample libraries or NotePerformer for the process of creation but only the final render when you either have to deliver a score to a conductor or I suppose a client (when the budget isn’t there for real musicians and hey let’s be honest, this latter option has been happening since the ‘80s with technology).
Click to expand...
This sounds like very good idea. How can composers and producers get their best possible rendition of their work, without having the inconvience of booking, rehearsing and recording lots of sessions, yet the performers are still compensated. Near live recording level of realism are reserved for per case use that is. Or there could be a pool of solo instrument players tied to an ai product, every one can collect royalties through income generated per renditions. I believe cevio is already doing something similar to this.
 
hazenut said:
This sounds like very good idea. How can composers and producers get their best possible rendition of their work, without having the inconvience of booking, rehearsing and recording lots of sessions, yet the performers are still compensated. Near live recording level of realism are reserved for per case use that is. Or there could be a pool of solo instrument players tied to an ai product, every one can collect royalties through income generated per renditions. I believe cevio is already doing something similar to this.
Click to expand...
I am all for compensating performers, as I would like to be compensated for the music I write. I know there is a lot of knee jerk assumptions that those of us who see the benefit of AI are somehow heretics. It's such a straw man position....
 
You must log in or register to reply here.
Top Bottom