"Most images in the dataset are monochrome, so the monochrome(greyscale) images will be more stable."

Checkpoint: waiNSFWIllustrious_v140

Lora: Noob Hyper Dmd; strength: 0.45-0.7 (Optional)

Lora: Simple/Transparent Background - Background (Optional)

Sampler name: euler | Scheduler: normal | Cfg: 1~1.3 | Steps: 5~8 (If you're not using Noob Hyper Dmd, please use a general metrics for generation)

HighRes-Fix: RealESRGAN_x4plus_anime_6B (upscale by 1.35, hires steps: 8~10, denoise: 0.1~0.3) (Optional, I didn't use HighRes in all samples)

Tool: ComfyUI

DISCLAIMER:
THIS LORA MODEL IS CREATED SOLELY FOR ACADEMIC RESEARCH AND PERSONAL USE. IT IS STRICTLY PROHIBITED TO USE IT FOR COMMERCIAL PURPOSES OR PROFIT. ADDITIONALLY, IT MUST NOT BE USED TO GENERATE NSFW CONTENT. ANY CONSEQUENCES ARISING FROM VIOLATIONS OF THESE TERMS SHALL BE THE SOLE RESPONSIBILITY OF THE USER.