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

MusicLM: Generating music from text

arxiv.org

Post reply on HN
Print view View original

51–60 of 114 posts

Re: MusicLM: Generating music from text

#51

Post by throw_pm23 »

echelon wrote: post #44
throwaway290 wrote: post #41

Earlier quoted context omitted.

I hope you also consider getting some music from an actual producer!

Why is there this need for humans to be involved at all stages of production? Do we still need humans to knit and handwash our clothes? To live in a caboose or lighthouse? To operate our elevators? To deliver the milk? Why should anyone have to learn to draw in the future when machines promise to do a better job? There are so many better uses for our time, and too many things to do for our short lives to handle. I wa…

It may be empowering, but it is not yet clear that the result will be appealing. Not trying to be an AI-skeptic, but I find art to still be mostly about communication; when I listen to some music I enjoy I feel like I have some shared experience with the author which they are able to communicate via music. I have yet to see this effect in AI-produced stuff, but even if the effect would be fully imitated, it is still not clear that it would have the same appeal.

An analogy: chess AI's are clearly superior to human chess by any reasonable measure. I still enjoy playing with other people (even online, even anonymously) infinitely more than playing with an AI, no matter how well calibrated/tuned to my level.

Re: MusicLM: Generating music from text

#53

Post by LunarAurora »

spyder wrote: post #37

Awesome, this is probably the best one to date even compared to the recent Riffusion, Jukebox and the older MIDI generating MuseNet. especially the conditioning on humming and whistling examples are cool, but to bad they use very common melodies for that so it's easier job for the model and harder for us to judge how well would it work on less common melodies.

> especially the conditioning on humming and whistling ... but to bad they use very common melodies

100%. Now if you add the detailed "Painting Caption Conditioning" to the mix, you can create a melody in the style of… an image, which, in a way, is kind of a controlled artificial synaesthesia [1].

[1] https://en.wikipedia.org/wiki/Synesthesia

Re: MusicLM: Generating music from text

#54

Post by b0afc375b5 »

highhedgehog wrote: post #35
snowram wrote: post #32

Earlier quoted context omitted.

May I recommend to learn how to use a tracker software? It is very fun to play with!

what's a tracker software?

https://youtu.be/roBkg-iPrbw

I think this showed up on my recommendation. I didn't know anything about tracker music, but this video explains it chronologically and very well.

Re: MusicLM: Generating music from text

#55

Post by adriand »

modeless wrote: post #13
georgehill wrote: post #2

Demo: https://google-research.github.io/seanet/musiclm/examples/

I love that there's a whole section for accordion examples. I think the accordion rap has a bad word in it. Accordion techno works surprisingly well. They really need to bump up to 48 KHz so all the music doesn't sound like it's being played over a telephone. A factor of two in the cost shouldn't be prohibitive. So much of the audio generation stuff I've seen has fatal flaws like this baked into the dataset and/or tr…

This is the best music that incorporates AI that I have heard by a long shot - process is here:

https://ooo.ghostbows.ooo/about/

Music is on all the usual streaming platforms, the album is called Shadow Planet, band is The Cotton Modules. Also avail on their website:

https://ooo.ghostbows.ooo/

Re: MusicLM: Generating music from text

#56

Post by ben_w »

FeepingCreature wrote: post #23

Weights when though. :-(

For this specific paper, probably never. From the final section:

"""We acknowledge the risk of potential misappropriation of creative content associated to the use-case. In accordance with responsible model development practices, we conducted a thorough study of memorization, adapting and extending a methodology used in the context of text-based LLMs, focusing on the semantic modeling stage. We found that only a tiny fraction of examples was memorized exactly, while for 1% of the examples we could identify an approximate match. We strongly emphasize the need for more future work in tackling these risks associated to music generation — we have no plans to release models at this point."""

Re: MusicLM: Generating music from text

#57

Post by notahacker »

infinitifall wrote: post #48
dariosalvi78 wrote: post #45

AI music is the next thing coming, wait for copyright lawsuits to fall like bombs. I find it fascinating that MidJourney can make a 3D model of my face from a low quality image, rotate it in space, apply it on someone's else body, add coherent shadows and backgrounds, with a very credible result, and yet an AI cannot generate a decent song, which is 1-dimensional and has probably much less internal modelling to care…

Possibly just anecdotal evidence, but it seems like the pool of good music composers is much much smaller than the pool of good visual artists. Maybe that is an indicator of how "hard" each field is.

Think that depends heavily on what you consider to be a "composer" and what you consider to be "good" in both art and music.

Writing somewhat novel music to a formula arguably has a much lower skill bar than producing good representative art (especially if you allow sequencers as composition and disallow basic digital retouching of photos as visual art), but producing something that genuinely stands out may be harder

Re: MusicLM: Generating music from text

#58

Post by jtode »

dariosalvi78 wrote: post #45

AI music is the next thing coming, wait for copyright lawsuits to fall like bombs. I find it fascinating that MidJourney can make a 3D model of my face from a low quality image, rotate it in space, apply it on someone's else body, add coherent shadows and backgrounds, with a very credible result, and yet an AI cannot generate a decent song, which is 1-dimensional and has probably much less internal modelling to care…

It will be as good as the music it's trained on, no better and probably no worse.

Re: MusicLM: Generating music from text

#59

Post by jtode »

echelon wrote: post #44
throwaway290 wrote: post #41

Earlier quoted context omitted.

I hope you also consider getting some music from an actual producer!

Why is there this need for humans to be involved at all stages of production? Do we still need humans to knit and handwash our clothes? To live in a caboose or lighthouse? To operate our elevators? To deliver the milk? Why should anyone have to learn to draw in the future when machines promise to do a better job? There are so many better uses for our time, and too many things to do for our short lives to handle. I wa…

These AIs will only ever be as good as what gets fed into them. It follows that without something going in, they will output nothing. That source material will always be human in nature, at least until these neural networks get large enough for emergent consciousness to exist, at which point we might see actual creativity from a machine.

An AI without new inputs to consume will not evolve creatively, and you will get bored of its output quite quickly, I think.

Re: MusicLM: Generating music from text

#60

Post by LukaszWiktor »

oidar wrote: post #9
winrid wrote: post #6

Earlier quoted context omitted.

The vocals (on second page of table) are really interesting. If you walked into a club, you might not notice they are nonsense, but it's just good enough BS to be convincing.

I just popped in to mention that part. It kinda feels like when Dalle generates text - like from a distance you recognize it, but it is utterly gibberish. It's interesting because there are certain parts of machine generated outputs we zero in on - for image generation it's typically eyes and, if it has it, text. It will be interesting to watch what evolves as tell-tale sign for computer generated things.

Eyes or rather hands?
Post reply on HN