yes, new model. nothing to do with stable diffusion

Yes, but it doesn't have the usual SDXL workflow. Go on huggingface.co to see the model card and how it's used

for anyone reading this - dont waste time - uses more resources and gives terrible output for that kind of resources and does not have any kind of optimizations available. not worth it

I believe that was for training, not inference

@condzero1950 You're right. I've been using PixArt Sigma alongside Photon in a comfyui workflow, and it's been consuming around 10GB of RAM and 6 to 8GB of VRAM.

and still not worthy

@aiaicaptain It's hit or miss.

Try the workflow in this image and let me know what you think?

https://civitai.com/images/16016846

@admajic My attempts did not come out as well as yours. I tried a few different samplers. The biggest difference I noticed is the eyes in your's are much more defined. Not sure why mine didn't come out better but I didn't run batches and cherry pick.

@condzero1950 did you download the png and run it into comfyui? Try that and let me know how you go

@admajic I am running on my own local machine, no online used.

@condzero1950 look up comfyui. It's run locally.

in comfyui - yes - same as sdxl - no