Flux 1 Dev Schnell On ComfyUI Local Installation Guide - The Best Image Model Of The Year?

Flux 1 The Best AI Image Model So Far? Better Than SD and Midjourney?

We delve into the revolutionary FLUX.1 model suite by Black Forest Labs, a cutting-edge advancement in generative AI. Explore the state-of-the-art text-to-image models, the FLUX 1 variants, and their exceptional capabilities in image generation. Witness how these models redefine creativity, efficiency, and diversity in AI-generated media, setting new standards in visual quality and output diversity.

For all FLUX.1 Models, Demo Page, ComfyUI Instruction:

Discover the technical aspects of FLUX models, their impressive 12 billion parameters, and advanced techniques that enhance performance and efficiency. Learn about the different variants, including FLUX 1 pro, FLUX 1 dev, and FLUX 1 schnell, each catering to specific needs and applications.

Furthermore, uncover the process of running FLUX in ComfyUI, from downloading the necessary model files to setting up the environment for seamless operation. Get insights into the hardware requirements, model compatibility, and the benefits of using FLUX models for text-to-image workflows.

Join us on this exciting journey through the world of generative AI and witness the transformative capabilities of the FLUX.1 model suite. Stay tuned for more updates, demonstrations, and insights into the latest advancements in AI technology.

If You Like tutorial like this, You Can Support Our Work In Patreon:

[TheFutureThinker]

For all FLUX.1 Models, Demo Page, ComfyUI Instruction:

[nbkxqnp]

why is my "Load Diffusion model" is not loading on my ComfyUI, any help ?

[mexihcahcoatl]

3080 TI 8vram and 16 Ram 1024 at 3 minutes using lowvram settings great images

[DevoutJourney]

when i pull the bottle image from comfyUI to load the work flow it doesnt give me these nodes. I get load checkpoint and some other stuff. everything is where it should be, what am i doing wrong

[beetwing]

Yeah been playing with it since yesterday, it's day and night with Stable Diffusion. Can't wait to also see it running in Automatic1111, I think it easier to manipulate control net and Loras.

[SubhamMondal]

"Load Diffusion model" is not loading on my ComfyUI, even though I've updated everything.

[hleet]

the 8b version of flux scnell (non official) works with 8GB of vram 🎉

[hoccachhoc]

Thank you for making the tutorial video. This is my first time installing ComfyUI to test Flux1. I followed all the instructions and installed everything, but when I click on 'run_cpu.bat' (or 'nvidia'), sometimes the ComfyUI screen appears, and sometimes it doesn't. When it doesn't appear, the last command line shown is as below, and when I press any key, it just disappears and nothing happens. I've tried many times but still can't open it. I don't know what this error is. Could you please guide me on how to fix it? Thank you very much.


Press any key to continue..."

[Lightof-Truth]

Thanks! Installing now. I will try it on Runpod also.

[One_Harmony]

So, amd does not auto work for windowd? And to make amd gpu work, it must be in a linux environment? Help us detail a tutorial to use comfyui work specifically with AMD gpu. I have msi armor radeon rx580 8g. Thanks. Looking forward for your video.

[marcoantonionunezcosinga]

Greetings, thank you. I hope soon there will be a "FLUX" model that does not require many PC resources. Now only the "lightning" model remains

[LastEsper]

Strange... I'm getting pure noise output (yes I've confirmed all the models are in the right places and Comfy is updated). On a M2 Max machine with 64GM RAM no less.

[kdstudio]

is it working with low ram like 6gb vram ??? with comfyUI

[TubeTreat]

7:15 how to get that node? Flux Guidance?

[szach-i-mat]

Im very upset. When i used my favorites sentese for every new model:

a hedgehog in a yellow helmet rides a red bike down the street in a large amusement park

He genered me SONIC cartoon and bike with name Sonic 😥😱😐 very bad 😐

[manchumuq]

My AMD 5950X + NVIDIA RTX3090 rig needs 260 seconds to generate a normal 1024 x 1024 image, and if I turn off the CUDA fallback policy, it would just show insufficient memory, so I guess this is exclusive to RTX 4090 and up, not just any card with 24GB of VRAM. I'll try the fp8 version now.

EDIT: I switched the weight_dtype to those two fp8, now it works like a charm, much slower than normal SDXL, but 1.3 - 1.6 s/it is already practical, considering it's basically one shot, no need to do much of inpaint or upscale or anything.

[crazyleafdesignweb]

this model look promising, as a base model it can generate image like this.
So all your images are generate from 1 sampler?

[-Belshazzar-]

This model is amazing indeed, but if I'll neat pick a bit, for me the only downside (except the lack of negative prompt which I am sure will change) is the fact it adds way too much bokeh,
I mean the bokeh is beautifully rendered, but I haven't managed to make even one image without this heavy bokeh, so essentially if you have a main subject, the background is always blurred

[dfgoijoi]

the file name extension is now ae.safetensors is it correct ? does it make any difference ? thanks

[SFzip]

Thanks! Installing now. Does it support multi-GPUs?