For best outputs: Sampler: Euler a / DPM++ 2M Karras Steps: 20 ~ 40, cfg 5 ~ 7, CLIP skip 2. VAE: kl-f8-anime2_fp16.safetensors. Pos prompt: This model usually doesn't require of big weights, but it's recommended to use the LoRa "FaceBeauty_qinglong_V3" with a weight of 1 for small faces like full body photos. Neg prompt: usually u can leave it empty, or reach some more of realism with the embedding "epiCPhoto-neg" included in the example runs.