Aitrepreneur
patreon
FLUX - THE BEST IMAGE AI MODEL EVER MADE!
🕑 Added 2024-08-03 17:09:53 +0000 UTCHey everyone! I've just released a new video on how to install the newest AI model tool called "Flux" that allows you to generate any images you want from text, exactly like Stable Diffusion or Midjourney or Dalle3 but just like Stable Diffusion it is free and open source. However the quality of the generation is leaps and bounds compared to any other models, it's just insane, amazing quality and the generation follows the prompt extremely well, making it the most powerful and flexible image AI model there is right now, period, it’s just fantastic.
You can check out the video right here: https://youtu.be/stOiAuyVnyQ
I also created a 1-click installer for the Webui for my Patreon supporters for Local AND Runpod, that you can find right here: https://www.patreon.com/posts/109398241
Down below you will also find the comfyui workflow for free for you to use, for both image generation and Image to Image!
As always, supporting me on Patreon allows me to keep creating helpful resources like this for the community. Thank you for your support - now go have some fun😉!
Comments
Aitrepreneur
use the latest workflow and installer I posted on Patreon instead
Marie-Anne Schmitz
SamplerCustomAdvanced mat1 and mat2 shapes cannot be multiplied (1x1280 and 768x3072)
Marie-Anne Schmitz
Hi ! First thanks for all ! i got this error while trying to generate an image : RuntimeError: mat1 and mat2 shapes cannot be multiplied (1x1280 and 768x3072)
Federico Bicocchi
second best. the best offline ai gen has boobs. flux has not boobs.
Aitrepreneur
try this one: https://we.tl/t-HN4NRcozO7
Ribez665@hotmail.co.uk
Couldn't get this to run, it just crashes; set it in a new CMD window to record the process. Looks like it can't download the ComfyUI file or find it - can I do this manually, or is there a simpler fix? Enter your choice (A or B) and press Enter: a Do you want to download the FLUX SCHNELL Model? A) YES B) NO Enter your choice (A or B) and press Enter: a Git already installed. Downloading ComfyUI... % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 0 0 0 0 0 0 0 0 --:--:-- --:--:-- --:--:-- 0 0 0 0 0 0 0 0 0 --:--:-- --:--:-- --:--:-- 0 curl: (35) schannel: next InitializeSecurityContext failed: CRYPT_E_NO_REVOCATION_CHECK (0x80092012) - The revocation function was unable to check revocation for the certificate. Extracting ComfyUI... 7-Zip 24.04 (x64) : Copyright (c) 1999-2024 Igor Pavlov : 2024-04-05 Scanning the drive for archives: ERROR: The system cannot find the file specified. ComfyUI_windows_portable_nvidia_cu121_or_cpu.7z System ERROR: The system cannot find the file specified. Extraction failed. Please check the downloaded file and try again. Running on NVIDA4090 32GB RAM
Aitrepreneur
The workflow json files are in the patreon posts, just load the 01 json file
Spawngts
Sorry if this is a dumb question. But where do I find workflows? I tried the Default Workflow and I get an error message when I try to Queue Prompt. It shows "flux1-dev-fp8.safetensors" in the "Load Checkpoint" box. I can't get it to produce an image. But, if I select "v1-5-pruned.ckpt", (which I manually loaded) it will load an image. But that's not a FLUX checkpoint, right?
Aitrepreneur
No idea, there is no info about this model. It's probably the same the one we use but done with a different configuration? I don't think there is any reason to use this one since the one we use right now has no quality difference over the normal model anyway
MrPittStone
https://huggingface.co/Comfy-Org/flux1-dev/tree/main there is it with 16GB. Is some merged stuff in this?
Aitrepreneur
you don't need to put any files anywhere manually, everything is done automatically. Just update comfyui and load a workflow
Aitrepreneur
I don't see any 16gb one, where do you see that?
Spawngts
Where do you put the .safetensors file? I put it in ComfyUI > models > checkpoints....but I get an error when I try to process an image.
MrPittStone
The Automatic Downloaded "unet" model "flux1-dev-fp8.safetensors" is 11GB. But on huggingface.co/Kijai/flux-fp8 ther is a "flux1-dev-fp8.safetensors" with 16GB. Whats the Difference?
RT Solvalou
I assumed as much. I am excited at how easily it responded to some of the prompts I gave it. But I guess I will have to rely on photoshop for now to give it the finishing touches that I want. At least I am no longer fixing hands and eyeballs and instead just replacing uniforms and starships.
Aitrepreneur
Haha man that's not how this works at all, people are still trying to see how and if that model can be trained correctly and second of all you can use comfyui to train a lora but it's not compatible with flux right now, you need to be patient the model came out like a few days ago, it's still super new
RT Solvalou
Cool. I will try that. The other question is, since I am a Star Trek fan, I would like to train the model on what different kinds of Star Trek ships there are. I was wondering if there is a way that I can configure the comfyUI so that I can add training function? If you did a video on ComfyUI to show how to configure it to run training prompts, can you link it? Thanks for your hard work and support.
Aitrepreneur
Flux doesn't support negative prompts but you can "hack it" with a special workflow here: https://we.tl/t-OLS3DFWQWg First update comfyui. Load the 03 workflow json file, then go to manager, click on "Install missing Nodes manager", there you will see 2 entries: Dynamic Thresholding Extra Models for ComfyUI Click on the checkbox to select the 2 entries then click install then click restart to restart the webui. Then you can write a prompt, add negative prompts
RT Solvalou
Working flawlessly on my 4070. Are we correct in assuming there is no negative prompting yet for this model? I am having a hard time only providing positive prompts.
