This model is a LoRA designed to repair the acceleration capability of Z-Image Turbo LoRA.
LoRAs directly trained based on Z-Image Turbo lose their acceleration ability, resulting in blurry images when generated under accelerated settings (steps=8, cfg=1), while the images generated under non-accelerated settings (steps=30, cfg=2) are normal.
DiffSynth-Studio/Z-Image-Turbo-DistillPatch
Submitted 1 week ago by Even_Adder@lemmy.dbzer0.com to stable_diffusion@lemmy.dbzer0.com
https://modelscope.cn/models/DiffSynth-Studio/Z-Image-Turbo-DistillPatch