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

101–110 of 159 posts

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

#101

Post by skort »

vessenes wrote: post #82
skort wrote: post #75

Earlier quoted context omitted.

>but, it's a real loss. Both in terms of human knowledge of, say composition, emotion, and so on, but also for style diversity But that real art still exists, and can still be found, so what exactly is the loss here?

We may differ on our take about the usefulness of diffusion models, but I'd say it's a loss in that many of the visuals humans will see in the next ten years are going to be generated by these models, and I for one wish they weren't just trained on weeb shit.

You'll still be able to ask a person to create art in a specific style if you'd like.

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

#102

Post by GaggiX »

SV_BubbleTime wrote: post #100
GaggiX wrote: post #97

Earlier quoted context omitted.

Have you run Flux Pro offline?

No, only a dozen Flux Dev models different distillations, quantizations, and fine-tunes with LORAs. But you keep pretending that close source AI is a sustainable comparison.

Flux Pro (v1 and v1.1) is a close source model.

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

#104

Post by drdaeman »

morbicer wrote: post #93
drdaeman wrote: post #56

Earlier quoted context omitted.

Yet, it doesn't seem to know how a Tektronix 4010 actually looks like... ;) I had similar issues trying to paint a "I cast non-magic missile" meme with a fantasy wizard using a missile launcher. No model out there (I've tried SD, SDXL, FLUX.1dev and now this FLUX1.1pro) knows how a missile launcher looks like (neither as a generic term, nor any specific systems) and even has no clue how it's held, so they all draw re…

Isn't it because the shoulder launched weapon is usually called rocket launcher, rpg or bazooka? Never heard it referred as misille launcher.

I've tried all of those and then some (e.g. "ATGM"), plus various specific names (like "FGM-148 Javelin", "M1 Bazooka", or "RPG-7", which are all quite iconic and well-recognized so I thought some of those may appear in training data) - all no bueno. Models are simply unaware about such devices, best of their "guesses" is that it's a weapon, so they draw something rifle- or pistol-shaped.

And, sure, that's what LoRAs are for. If I can figure out how to train one for FLUX, in a way that would actually produce something meaningful (my pitiful attempts at SDXL LoRA training were... less that stellar, and FLUX is quite different from everything). Although that's probably not worth it for making a meme picture...

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

#105

Post by ilaksh »

loxias wrote: post #94
ilaksh wrote: post #48

Pretty smart model. Here's one I made: https://replicate.com/p/6ez0x8xqvsrga0cjadg8m7bah0

It's quite good at following a detailed paragraph long description of an scene, which is a double edged sword. A lot of the fun for me with early text to image models was underspecifying an image and then enjoying how the model "invents" it. "Steampunk spaceship", "communist bear", "glass city". flux is amazing, but I find it requires a very literal description, which pushes the "creative work" back to the text itsel…

Prompt enhancement is now a standard feature in many image generation tools.

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

#106

Post by spython »

davidbarker wrote: post #87
vessenes wrote: post #80

Earlier quoted context omitted.

I like those, and there's an electroshock lora that's just awesome out there. That said, Tarot and others like it are "illustrator" type styles with extra juice. I have not successfully trained a LoRa for any painting style, Flux does not seem to know about painting.

I'm curious to give this a go. I've been training a lot of LoRAs for FLUX dev recently (purely for fun). I'm sure there must be a way to get this working. Here are a few I've recently trained: https://civitai.com/user/dvyio

This looks really good! What is your process to get this kind of high quality LoRAs?

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

#107

Post by LeoPanthera »

smcleod wrote: post #90
LeoPanthera wrote: post #70

Earlier quoted context omitted.

"Draw Things" is a native Mac app for text to image. It's a a lot more advanced than DiffusionBee, it will download the models for you, and it's free. It's also available for iOS. (!)

Draw things is neat but it's so damn slow compared to other tools (e.g. invokeai), I'm not sure why it takes so long to generate images with any model?

It's not any slower than invokeai for me. Maybe check the settings, and try using the GPU instead of CoreML.

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

#108

Post by liuliu »

smcleod wrote: post #90
LeoPanthera wrote: post #70

Earlier quoted context omitted.

"Draw Things" is a native Mac app for text to image. It's a a lot more advanced than DiffusionBee, it will download the models for you, and it's free. It's also available for iOS. (!)

Draw things is neat but it's so damn slow compared to other tools (e.g. invokeai), I'm not sure why it takes so long to generate images with any model?

On the same Mac hardware, Draw Things should be the fastest on models such as SDXL / FLUX.1 against other tools based on PyTorch (I stopped benchmarking SD v1.5 results for a while so that might regress a little bit here or there).

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

#109

Post by liuliu »

zamadatix wrote: post #66
drcongo wrote: post #60

Earlier quoted context omitted.

Really? I tried using it in ComfyUI on my Mac Studio, failed, went searching for answers and all I could find said that something something fp8 can't run on a Mac, so I moved on.

If you're looking for a prebuilt "no tinkering" solution https://diffusionbee.com/ is an open source app (Github link at the bottom of the page if you want to see the code) which has a built in button to import Flux models at the bottom of the home screen.

I usually don't want to comment on these, but: DiffusionBee's repo https://github.com/divamgupta/diffusionbee-stable-diffusion-... don't have any updates for 9 months except regular binary releases. There is no source code available for their recent builds. I think it is a bit unfair to say it is open-source app at this point given you are using a binary probably far different from the repo.

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

#110

Post by davidbarker »

vessenes wrote: post #88
vessenes wrote: post #80

Earlier quoted context omitted.

I like those, and there's an electroshock lora that's just awesome out there. That said, Tarot and others like it are "illustrator" type styles with extra juice. I have not successfully trained a LoRa for any painting style, Flux does not seem to know about painting.

@davidbarker -- please do, that sounds awesome! I did not have good results.

It's trickier than I thought it would be.

Here are a few in Degar style I made after training for 2,500 steps. I'd love to hear what you think of them. To my (untrained) eye, they seem a little too defined, perhaps?

https://imgur.com/a/sqsQLPg

Post reply on HN