Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Open-sourcing AudioCraft: Generative AI for audio

ai.meta.com

Post reply on HN
Print view View original

61–70 of 335 posts

Re: Open-sourcing AudioCraft: Generative AI for audio

#61

Post by dragonwriter »

analognoise wrote: post #39
eropple wrote: post #34

Earlier quoted context omitted.

> MusicGen's weights are licensed CC-BY-NC, which is effectively a nonlicense as there is no noncommercial use you could make of an art generator How do you figure? Have you never just...made stuff to make stuff?

I think the key word there is "noncommercial".

Yes, but you can easily make noncommercial use of an art generator.

Obviously, you can't host a commercial art generation service with a noncommercial-use license, and (insofar as art produced by a generator is a derivative work of the model weights, which is a controversial and untested legal theory) you can’t make commercial art with a noncommercial license, but not all art is commercial.

Re: Open-sourcing AudioCraft: Generative AI for audio

#62

Post by Tepix »

ericpauley wrote: post #43
kmeisthax wrote: post #27

Earlier quoted context omitted.

This is purely a function of everyone remembering the RIAA's decade-long campaign to prevent people from taking the music they had rightfully stolen. As far as I'm aware LLaMA was trained on "publicly available data"[0], not "licensed data". Furthermore, MusicGen's weights are licensed CC-BY-NC, which is effectively a nonlicense as there is no noncommercial use you could make of an art generator[1]. This is not only…

My understanding (IANAL) [1] is that copyright licenses have no say on the output of software. Further, CC licenses don't say anything about running or using software (or model weights). It's therefore questionable whether the CC-BY-NC license actually prevents commercial use of the model. [1] https://opensource.stackexchange.com/questions/12070/allowed...

You're wrong because software, as you describe it, includes the "cp" command which creates a perfect copy.

Re: Open-sourcing AudioCraft: Generative AI for audio

#64

Post by tikhonj »

Tepix wrote: post #62
ericpauley wrote: post #43

Earlier quoted context omitted.

My understanding (IANAL) [1] is that copyright licenses have no say on the output of software. Further, CC licenses don't say anything about running or using software (or model weights). It's therefore questionable whether the CC-BY-NC license actually prevents commercial use of the model. [1] https://opensource.stackexchange.com/questions/12070/allowed...

You're wrong because software, as you describe it, includes the "cp" command which creates a perfect copy.

The copyright license of the cp code itself has no bearing on the copyright of what you produce (well, copy) with cp.

Re: Open-sourcing AudioCraft: Generative AI for audio

#66

Post by reducesuffering »

"Now, increasingly, we live in a world where more and more of these cultural artifacts will be coming from an alien intelligence. Very quickly we might reach a point when most of the stories, images, songs, TV shows, whatever are created by an alien intelligence.

And if we now find ourselves inside this kind of world of illusions created by an alien intelligence that we don’t understand, but it understands us, this is a kind of spiritual enslavement that we won’t be able to break out of because it understands us. It understands how to manipulate us, but we don’t understand what is behind this screen of stories and images and songs."

-Yuval Noah Harari

Re: Open-sourcing AudioCraft: Generative AI for audio

#67

Post by croes »

gavman wrote: post #13

> MusicGen, which was trained with Meta-owned and specifically licensed music, generates music from text-based user inputs, while AudioGen, which was trained on public sound effects, generates audio from text-based user inputs. Meta is really clearly trying to differentiate themselves from OpenAI here. Open source + driving home "we don't use data we haven't paid for / don't own".

"If we don't win here, then at least we'll kick their lawn to pieces."

Re: Open-sourcing AudioCraft: Generative AI for audio

#68

Post by manquer »

PaulDavisThe1st wrote: post #53
kmeisthax wrote: post #31

Earlier quoted context omitted.

The RIAA pioneered copyright enforcement at the individual level back in the 2000s, they absolutely would try to sue downstream AudioCraft users.

[flagged]

Even before streaming, You never "owned" any music legally [1], you merely owned a physical copy of a performance[3] of a song, that in no way gives you the right to make derivative works [2] automatically.

Also it doesn't really matter on what the law says, RIAA in the last iteration, relied on the fact it you would rather pay a fine than be able to afford expensive lawyers to fight the specifics out in court on average.

It was always about disproportionate ability to bring resources against individual "offenders" to create fear among everyone to deter "undesirable" forms of copying, not necessarily what the legal protections were.

---

[1] Unless you specifically commissioned it under a contract which gave you the right

[2] See recent cases including those related to Kris Kashtanova and Andy Warhol.

[3] Not the song, just the performance, aka Taylor Swift version, for good explanation of how the rights are divvied up in the music industry a Planet Money series covers it well https://www.npr.org/sections/money/2022/10/29/1131927591/inf...

Re: Open-sourcing AudioCraft: Generative AI for audio

#70

Post by Blackthorn »

kmeisthax wrote: post #27
gavman wrote: post #13

> MusicGen, which was trained with Meta-owned and specifically licensed music, generates music from text-based user inputs, while AudioGen, which was trained on public sound effects, generates audio from text-based user inputs. Meta is really clearly trying to differentiate themselves from OpenAI here. Open source + driving home "we don't use data we haven't paid for / don't own".

This is purely a function of everyone remembering the RIAA's decade-long campaign to prevent people from taking the music they had rightfully stolen. As far as I'm aware LLaMA was trained on "publicly available data"[0], not "licensed data". Furthermore, MusicGen's weights are licensed CC-BY-NC, which is effectively a nonlicense as there is no noncommercial use you could make of an art generator[1]. This is not only…

> there is no noncommercial use you could make of an art generator

I'm sorry, what?

Post reply on HN