Information
This is an experimental LoRA. It gives mixed results.
I have found it to 'smooth/refine' Z-Image generations.
Please share any results or issues!!
Use
No trigger word needed.
Use recommended settings.
CFG 2-3, steps 20-30
Example generations done with: res_multistep/simple/3.0 cfg/18 steps
Training
Base Model: Z-Image-De-Turbo
Dataset: Echo-4o-Image