Updated: Feb 26, 2025
styleVerified: a month ago
SafeTensor
Type | |
Stats | 480 3,924 817 |
Reviews | (118) |
Published | Feb 26, 2025 |
Base Model | |
Trigger Words | no lineart |
Hash | AutoV2 88146B0985 |
u-net only lora trained on Takarai Shunsuke for noob eps 0.5.
Using the "no lineart" tag isn't necessary but you probably want to use it, anyway.
Like my kiato lora, this lora also works fairly well as a drop-in style replacement with high denoise i2i upscaling. Mostly mentioning this since there might be odd/conflicting anatomy to what you may want when using it alone as a base style.
To which; recommended i2i upscale settings:
xinsir controlnet tile, .7-.8 denoise, tile_resample for preprocessor, .35 control weight and .45 end step