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

FLUX1.1 [pro] – New SotA text-to-image model from Black Forest Labs

replicate.com

Post reply on HN
Print view View original

21–30 of 159 posts

Re: FLUX1.1 [pro] – New SotA text-to-image model from Black Forest Labs

#21

Post by waffletower »

minimaxir wrote: post #10
nirav72 wrote: post #8

Are there any projects that allow for easy setup and hosting Flux locally? Similar to SD projects like InvokeAI or a1111

Flux is more weird than old SD projects since Flux is extremely resource dependant and won't run on most hardware.

Doesn't take a lot of effort to get Flux dev/schnell to run on 3090s unquantized, but I agree that 24gb is the consumer GPU memory limit and there are many with less than that. Flux runs great on modern Mac hardware as well, if you have at least 32gb of unified memory.

Re: FLUX1.1 [pro] – New SotA text-to-image model from Black Forest Labs

#22

Post by halJordan »

sharkjacobs wrote: post #6

"state of the art" has become such tired marketing jargon. "our most advanced and efficient model yet" "a significant step forward in our mission to empower creators" I get it, you can't sell things if you don't market them, and you can't make a living making things if you don't sell them, but it's exhausting.

It is state of the art. And it's not like the art has stagnated.

Re: FLUX1.1 [pro] – New SotA text-to-image model from Black Forest Labs

#23

Post by sharkjacobs »

minimaxir wrote: post #11
sharkjacobs wrote: post #6

"state of the art" has become such tired marketing jargon. "our most advanced and efficient model yet" "a significant step forward in our mission to empower creators" I get it, you can't sell things if you don't market them, and you can't make a living making things if you don't sell them, but it's exhausting.

The official blog post justifies the marketing copy a bit more with metrics.

The point is that the metrics say the thing, this stuff doesn't say actually anything.

What does "state of the art" mean? That it's using the latest "cutting edge" model technology?

When Apple releases a new iPhone Pro Max, it's "state of the art". When they release a new iPhone SE, there's an argument to be made that it's not because it uses 2 year old chips. But what would it even mean for BFL to release a model which wasn't "state of the art"

> our most advanced and efficient model yet

Yes, likewise, this is how technology companies work. They release something and then the next thing they release is more advanced.

> a significant step forward in our mission to empower creators

Going from 12 seconds to 4 seconds is a significant speed boost, but does it move the needle on their mission to empower creators? These are their words, not mine, it's a technical achievement and impressive incremental progress, but are there users out there who are more empowered by this? significantly more empowered!?

Re: FLUX1.1 [pro] – New SotA text-to-image model from Black Forest Labs

#25

Post by tripplyons »

Filligree wrote: post #14
minimaxir wrote: post #10

Earlier quoted context omitted.

Flux is more weird than old SD projects since Flux is extremely resource dependant and won't run on most hardware.

The GGUF quantisations do run on most recent hardware, albeit at increasingly concerning quality tradeoffs.

I haven't noticed any quality degradation with the 8-bit GGUF for Flux Dev, but I'm sure the smaller quantizations perform worse.

Re: FLUX1.1 [pro] – New SotA text-to-image model from Black Forest Labs

#26

Post by arizen »

sharkjacobs wrote: post #6

"state of the art" has become such tired marketing jargon. "our most advanced and efficient model yet" "a significant step forward in our mission to empower creators" I get it, you can't sell things if you don't market them, and you can't make a living making things if you don't sell them, but it's exhausting.

- How do copywriters greet each other in the morning?

- Take your morning to the next level!

Re: FLUX1.1 [pro] – New SotA text-to-image model from Black Forest Labs

#27

Post by throwaway314155 »

sharkjacobs wrote: post #23
minimaxir wrote: post #11

Earlier quoted context omitted.

The official blog post justifies the marketing copy a bit more with metrics.

The point is that the metrics say the thing, this stuff doesn't say actually anything. What does "state of the art" mean? That it's using the latest "cutting edge" model technology? When Apple releases a new iPhone Pro Max, it's "state of the art". When they release a new iPhone SE, there's an argument to be made that it's not because it uses 2 year old chips. But what would it even mean for BFL to release a model wh…

Holy shit the level of pedantry. State of the art in this context means it out performs all other models to date on standard evaluations, which is precisely what it does.

Did you miss the first flux release? Black forest labs aren't screwing around. The team consists of many of the _actual_ originators of Stable Diffusion's research (which was effectively co-opted by Emad Mostaque who is likely a sociopath).

Re: FLUX1.1 [pro] – New SotA text-to-image model from Black Forest Labs

#28

Post by vessenes »

Flux is so frustrating to me. Really good prompt adherence, strong ability to keep track of multiple parts of a scene, it's technically very impressive. However it seems to have had no training on art-art. I can't get it to generate even something that looks like Degas, for instance. And, I can't even fine tune a painterly art style of any sort into Flux dev. I get that there was working, living artist backlash at SD and I can therefore imagine that the BFL team has decided not to train on art, but, it's a real loss. Both in terms of human knowledge of, say composition, emotion, and so on, but also for style diversity.

For goodness sake, the MET in New York has a massive trove of open CC0 type licensed art. Dear BFL, please ease up a bit on this, and add some art-art to your models, they will be better as a result.

Re: FLUX1.1 [pro] – New SotA text-to-image model from Black Forest Labs

#29

Post by doctorpangloss »

nirav72 wrote: post #8

Are there any projects that allow for easy setup and hosting Flux locally? Similar to SD projects like InvokeAI or a1111

https://huggingface.co/docs/diffusers/main/en/api/pipelines/...

It's about 6 lines of Python.

Re: FLUX1.1 [pro] – New SotA text-to-image model from Black Forest Labs

#30

Post by GaggiX »

bemmu wrote: post #13
sharkjacobs wrote: post #6

"state of the art" has become such tired marketing jargon. "our most advanced and efficient model yet" "a significant step forward in our mission to empower creators" I get it, you can't sell things if you don't market them, and you can't make a living making things if you don't sell them, but it's exhausting.

Flux genuinely is the best model I’ve tried though. If there is a better one I’d love to know.

Have you tried Ideogram v2?
Post reply on HN