4-step distillation of Flux.2 now available

https://preview.redd.it/onnlnaws8v7g1.png?width=1024&format=png&auto=webp&s=bb8d210a6af676a317c8b9933446000d28cd74e7

Custom nodes: https://github.com/Lakonik/ComfyUI-piFlow?tab=readme-ov-file#pi-flux2
Model: https://huggingface.co/Lakonik/pi-FLUX.2
Demo: https://huggingface.co/spaces/Lakonik/pi-FLUX.2

Not sure if people are still interested in Flux.2, but here it is. Supports both text-to-image generation and multi-image editing in 4 or more steps.

https://redd.it/1ppedq9
@rStableDiffusion

Another method for increased Z-Image Seed Diversity

https://redd.it/1ppcj43
@rStableDiffusion

It's getting hot : PR for Z-Image Omni Base
https://redd.it/1ppjzdo
@rStableDiffusion

I made this Prompt-Builder for Z-Image/Flux/Nano-Banana

https://redd.it/1ppouf5
@rStableDiffusion

LMstudio with Qwen3 VL 8b and Z image turbo is the best combination

Using an already existing image in LMstudio with Qwen VL running and an enlarged context window with the prompt
"From what you see in the image, write me a detailed prompt for the AI ​​image generator, segment the prompt into subject, scene, style,..."
Use that prompt in ZIT and steps 10-20, and CFG 1 - 2 gives the best results depending on what you need.

https://redd.it/1pppf34
@rStableDiffusion

Z-Image is now the default image model on HuggingChat

https://redd.it/1ppq6nb
@rStableDiffusion

PSA It is pronounced "oiler"

Too many videos online mispronouncing the word when talking about using the euler scheduler. If you didn't know ~now you do~. "Oiler". I did the same thing when I read his name first learning, but PLEASE from now on, get it right!

https://redd.it/1ppsn77
@rStableDiffusion

ComfyUI Tutorial Series Ep 73: Final Episode & Z-Image ControlNet 2.0
https://www.youtube.com/watch?v=DMbXB6g17IU

https://redd.it/1ppu9hi
@rStableDiffusion