FLUX 1.1 - THIS look TOO REAL 😱 

oh boy. me and my swiss chees brain XD

Closest is api use

They do have a dev version of this model. I read an article about yesterday. I guess they would market the PRo 1.1 first for a few days before releasing the new dev 1.1

They need to make some money or it will end

@@holdthetruthhostage I have no issues with them making money, however a lot of the testing and development of workflows will be done by open source users. The biggest advertisers of the product are not usually those who are paying customers because they have every incentive to keep their workflows private.

Turkey doctors magicians

My skin is only touched by water and AI ;)

that's easy. You take a cat, add a beard and say the magic words "that's a very nice cat beard" :)

yes and no. even with anatmical correct illustrations, a hand has even in real life 5 or 3 or 7 or 2 fingers, depending on how you hold your hands. so it's pretty hard for AI to figure out what is what

@@OlivioSarikas it's the lack of anatomy training, see artists are trained to illustrate hands from literally cylinders & circles. AI is just feed images of hands, so how can it ever understand. Basically it's like showing you millions of cars without ever showing how it's built from scratch, sure you'll know how cars look and be able to replicate how they look but since you don't know how to make a car from scratch, how it's parts work you will struggle to create accurate car illustrations at all times that make anatomical & shape theory properly

@@holdthetruthhostage Ai doesn't learn like humans do. And it's not just hands. It's only more apparent with hands. AI can't create anything well if you actually test it. Try to create a face that is yawning in a realisting way. Try to create a perfect Ferrari. Try to create a checkered shirt. AI can't do it. What you think AI is doing, is looking at the form and understanding the form. But that is not at all what AI is doing. The diffusion AI we have right now only tries to turn noise into an image, based on where what pixel lands. It's a statistical tool. It has no understanding of the content. If you think AI understands an content, because image interrogation can "tell you what's in the image", then that is wrong. It only associates a bunch of pixels that look like a face with the word face. It would just as well link it to the word butt. Because ai does not know what a face is. it only compares values to other values and gives you the best match.

@@OlivioSarikas Thank you sir! you are a AI channel wich i normally dont like but at least you are not spreading wrong information on what transformer architecture actually is.

I'm sexy and I know it ;)

looks like ai-generated

good idea. Maybe we do that tomorrow in the stream :)

smoebody got to pay for the trainging ;)

the pro version is much larger and does not run local. just like you can't fit a 10.000 pages book into 500 pages. you can't fit a much more capable model into a tiny local size. As you can see with video games, things are getting larger, not smaller

@@OlivioSarikas Actually that is not correct. All Flux models have the same size: "All public FLUX.1 models are based on a hybrid architecture of multimodal and parallel diffusion transformer blocks and scaled to 12B parameters. We improve over previous state-of-the-art diffusion models by building on flow matching, a general and conceptually simple method for training generative models, which includes diffusion as a special case. In addition, we increase model performance and improve hardware efficiency by incorporating rotary positional embeddings and parallel attention layers." If we can run flux dev, we can run the pro. The difference is the level of distillation.

@@abdelhakkhalil7684 your text is about all public models. the pro model is not public. alsothere are flux models of different sizes from other sources like flux1-dev-F16.gguf has 22.3 GB, while the flux1-dev-Q2_K.gguf has only 4.03 GB and so on

@@OlivioSarikas Yeah but those are not officially launched from The Black Forest Lab. Those are quantized versions of the Dev and Schnell models. I just read that all three full precision variants have the same size. In their official website, the lab says this: "FLUX.1 [dev] is an open-weight, guidance-distilled model for non-commercial applications. Directly distilled from FLUX.1 [pro], FLUX.1 [dev] obtains similar quality and prompt adherence capabilities, while being more efficient than a standard model of the same size." The dev is at least very close in size to the Pro version. But anyway, I trust your words as you may know more that I do. Thank you for your replies and keep up the great work.

The fact we are asking this question “if we can run” on a next gen flagship GPU shows that we are cooked

not sure, but i don't see why not

360?

shhhhhhh.....

Somehow Midjourney is the only service that people love to pay for. But, they are also very good.

Pro never meant "advanced". Pro is short for professional. If you get paid, you are a pro, no matter how good you are.

@@OlivioSarikas "Pro is Professional" what i said... i'm just referring to programs and stuff that includes a "pro plan". However, there's nothing "professional" about this model, it's "Enterprise" oriented imho.

@@z1mt0n1x2 it's good for commercial use and you can run it through a api. not sure what more you need from this

The great wave is a print by Hokusai, not a sculpture as shown in the image ;)

yes, i said that wrong in the video

maybe it's time to bring him back :)

you can always helps us out by training a skin lora. it doesn't even cost much

Thank you very much. Yes, sadly lots of entitled people think they deserve only the best and for free out of any ai model. But I try to ban the worst karens

​@@OlivioSarikas I agree with what was said. I use your channel updates as my AI news feed and it's really allowed me to stay up to date on new releases. Much thanks!

@@archael18 thank you :)