Photorealistic AI Generated Images

Lovely work, my friend. More ComfyUI wizardry?
Hi, friend, sorry for the late reply. Yes, used ComfyUI for. The models (like the Z-Image Turbo flavor I used) are what allow for natural language prompting. I tend to use that mostly to see which models respond better to it. Glad you've incorporated that into your methods! Always great to find new things with this stuff, it just keeps getting better and better. :D
 
I've just installed Krea2 to run locally. My rig isn't powerful enough to run the full model so I'm using the smaller variant, results are promising so far.
I’m currently using a modified KR2 checkpoint. There’s an FP8 version as well, which is about half the size of the BF16 one. I’ve tried both and honestly couldn’t see any noticeable difference in quality. There’s also an NVFP4 version optimized for 8GB of VRAM.
 
  • Like
Reactions: CoolKevin
A little off-topic, but I thought I’d share this here as well, just to give an idea of where AI video generation with integrated audio is at right now. And the impressive part: this was all generated locally with MiniMax. It’s pretty incredible that they managed to squeeze all of this into just 12GB of VRAM.

LINK :)