V2 updated:V2 utilizes a wider variety of images with different styles as training data, resulting in more diverse and enriched outcomes when generating content.V2采用更多不同风格的图片作为训练素材，使得模型抽卡结果更加丰富！

Grace Chen

<ul><li>Use any checkpoint based on SD1.5</li><li>its 768,768 image</li><li>turn on Hires.fix (or not, please experiment with it yourself)</li><li>Hires steps: 14-17, denoising:04-65 ish.. just try it yourself</li><li>CGE: 5-7</li><li>Use DPM++ 2M Karras or DPM++ SDE Karras</li><li>Use 20-ish sampling steps</li><li>Strength: 0.6-1</li><li>Clip Skip: 1-3</li><li>I suggest turn off the restore faces check mark for better results of this Lora</li></ul>She might look like someone from real life, but she is not what you think. She is a totally fictional character.

Files with SHA256: acd5f5e5e0ad44d82c44dc231ec711d3bc13e09e2212bc72da1d139e4436dfcc