For best outputs:

Sampler: DPM++ 2M

Steps: 50 ~ 60, cfg 7, CLIP skip 2.

VAE: kl-f8-anime2_fp16.safetensors.

Pos prompt: start your pos with the classical "realistic, masterpiece, best quality, cinematic lighting, natural shadow, highest detail, professional photography" ect, then add the lora "FaceBeauty_qinglong_V3" at the end, with a weight of 1 (included in the example runs)

Neg prompt: "girl, woman, messy hair, render, cartoon, cgi, render, illustration, painting, drawing, cgi, 3D, paintings, sketches, (worst quality:2), (low quality:2), UnrealisticDream, BadDream, bad feets, bad ass, (normal quality:2), (badhandv4:0.7 , (deformed, distorted, disfigured:1.3), (gray hair:1.3), (face out of frame), cropped, Unspeakable-Horrors-Composition-4v, dark hair, brown, bad fingers, distorded fingers, bad res, bad quality, error, low quality, poorly drawn, a photocopy, a 3D render, a renaissance painting, computer graphics, arabesque, polycount, (blurry:1.5), out of focus, easynegative", being the "easynegative", "BadDream" and "Unspeakable-Horrors-Composition-4v", embeddings included in the example runs.

Note: Despite to be a model trained to create men, it can draw females perfectly too.