For best outputs:

Sampler: DPM++ SDE.

Steps: 20, cfg 7, CLIP skip 2.

VAE: kl-f8-anime2_fp16.safetensors

Pos prompt: just what you want to see. This model is set to semi-realism, understands well and usually doesn't need big weights. Duplicated people may appear with a 768 x 768 resolution; if prompting doesn't help, change it to 512 x 768.

Neg prompt: "(NSFW), (nipples), (worst quality:1.5), (low quality:1.5), (normal quality:1.5), bad eye, (zombie, sketch, interlocked fingers, comic)".

Checkpoint Merge.