Open-sourcing AudioCraft: Generative AI for audio • Hacker News Board

Re: Open-sourcing AudioCraft: Generative AI for audio

#51

Post by operator-name » Wed, Aug 02, 2023, 4:44 PM UTC

jasonjmcghee wrote: post #26

The difference between MBD-EnCodec and EnCodec is pretty interesting. MBD variant sounds more like a professional studio recording, while the EnCodec feels like a richer sound. Curious if I’m alone in that. (At the bottom https://audiocraft.metademolab.com/musicgen.html ) For what it’s worth though, the voice based examples sound dramatically better with MBD https://audiocraft.metademolab.com/encodec.html

MBD definitely sounds like it was recorded in dead room, whereas plain EnCodec has been mixed but includes some artificial noise.

Re: Open-sourcing AudioCraft: Generative AI for audio

#52

Post by PaulDavisThe1st » Wed, Aug 02, 2023, 4:45 PM UTC

throwaway290 wrote: post #46

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.

They should start with AudioCraft itself, conceptually it's derivative work and it doesn't matter if it's "open source" or not. Try throwing in someone's sample in a song and publish it saying "no copyright infringement intended and I totally don't make any money from it"... If it becomes popular, see how long it stays up until DMCA takedown. And we know this dataset is already popular.

> and publish it

This is precisely the opposite of the context I was remarking on.

Re: Open-sourcing AudioCraft: Generative AI for audio

#53

Post by PaulDavisThe1st » Wed, Aug 02, 2023, 4:46 PM UTC

kmeisthax wrote: post #31

PaulDavisThe1st wrote: post #25

Earlier quoted context omitted.

They can't litigate a person doing this at home, and never redistributing. I suppose they might try, anyway.

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

[flagged]

Re: Open-sourcing AudioCraft: Generative AI for audio

#54

Post by roody15 » Wed, Aug 02, 2023, 4:46 PM UTC

zyang wrote: post #44

unnouinceput wrote: post #36

Earlier quoted context omitted.

The same move that Microsoft did back in 90's to kill Netscape. Make your product the one available to masses, next generation of users will be using your product.

I was just thinking how Google made Android free to check Microsoft. This is Meta checking Google.

Checking Google or OpenAI (or both?)

Re: Open-sourcing AudioCraft: Generative AI for audio

#55

Post by ipaddr » Wed, Aug 02, 2023, 4:46 PM UTC

zyang wrote: post #44

unnouinceput wrote: post #36

Earlier quoted context omitted.

The same move that Microsoft did back in 90's to kill Netscape. Make your product the one available to masses, next generation of users will be using your product.

I was just thinking how Google made Android free to check Microsoft. This is Meta checking Google.

Checking OpenAI. Google is still playing checkers.

Re: Open-sourcing AudioCraft: Generative AI for audio

#56

Post by sangnoir » Wed, Aug 02, 2023, 4:47 PM UTC

Culonavirus wrote: post #14

Here's a different question: Can you use the audio output this produces for anything else other than "research purposes"?

You can - as long as it's not commercial. It's a broad definition, but a good eule of thumb is if you're not directly making money out of the generated audio. They may still cone for you if you're making money indirectly, so consult a lawyer.

Re: Open-sourcing AudioCraft: Generative AI for audio

#57

Post by kypro » Wed, Aug 02, 2023, 4:48 PM UTC

s1k3s wrote: post #4

The demos are great. Could someone explain what’s in it for Meta open sourcing all these models?

A competitive opensource project basically destroys the pricing power of all closed-source alternatives.

If you're a company and wanted to integrate an LLM into your product if the choice is between several equally good models, but one is free and open-source which would you pick?

Aside from keeping competition at bay, this move also gives Meta leverage because ecosystems are now being built around their projects. If these models see wide-scale adoption they could later launch AudioCraft+ as a licensed version with some extra features for example.

Alternatively, they might offer support or hosting for their open source projects.

Right now though I think the primary benefit of these open sourced models is to attract talent. If Meta is seen as one of the leaders in AI then researchers will want to work for them simply for the prestige.

Arguably one of the reasons Meta has been behind so many awesome projects like PyTorch and React over the last decade was because they were seen as the cool place for recently graduated, but talented software engineers to work in ~2010.

Re: Open-sourcing AudioCraft: Generative AI for audio

#58

Post by ChildOfChaos » Wed, Aug 02, 2023, 4:50 PM UTC

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".

It's likely partly a PR/branding exercise as well.

In the new world that Meta sees, of VR/AR and AI, Meta is in a position already were people don't want them to have much power in this world, because they don't trust them over privacy etc, meta is trying to pivot to become more trustworthy so they make genuine moves in this space.

Re: Open-sourcing AudioCraft: Generative AI for audio

#59

Post by illwrks » Wed, Aug 02, 2023, 4:50 PM UTC

joshstrange wrote: post #48

I'm looking forward to playing with the M1 Mac apps/cli-tools that will probably come out for this in the next week or so! Being able to run this stuff locally is a lot of fun.

Are the M1 macs capable enough? I'm eyeing and upgrade in the coming months and I'm curious if a MacBook would be suitable

Re: Open-sourcing AudioCraft: Generative AI for audio

#60

Post by upshide » Wed, Aug 02, 2023, 4:51 PM UTC

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".

[dead]