Recommended Weight: 0.8

This model is currently the LoRA I've spent the most time on, used the most dataset materials for, offers the best generalization, and is my personal favorite. It delivers exceptional performance in Asian portraits, skin texture, scenes, and details across all aspects. It was trained using natural language captions, so it is recommended to write prompts using natural language descriptions. Whether for e-commerce, Xiaohongshu, or Douyin, it is highly suitable and a must-have LoRA. All sample images were generated directly online on Lib without any post-editing—what you see is what you get. I look forward to seeing your posted images!

CFG: 3.5

Steps: 35

Recommended Resolution: 896*1152