Flux.1-Lite-8B-alpha, 4 steps merge, FP8

custom merge with low-step loras. base model: https://huggingface.co/Freepik/flux.1-lite-8B-alpha

1080p picture with 4 steps takes 10 seconds on a 4070 ti

currently working only in comfyui via load diffusion model node.

works with some loras, but not all.

PuLID not working.