Description
Sampling method:DPM++ 2M Karras
Iterations:30
High-resolution restoration: Upscaling algorithm:R-ESRGAN 4x+, Repaint magnitude:0.4
Lora weight 0.8, Trigger word white yarn
Legend model:Wheat Orange_v7 General realistic large models can all be tried
Image size recommendation 512*768,
Turn on high-resolution restoration, turn on face repair plug-in, manual comparison is prone to crash, so it is better to draw more cards