For best outputs:

Sampler: DPM++ 2M Karras.

Steps: 20, cfg 7, CLIP skip 2.

VAE: kl-f8-anime2_fp16.safetensors

Pos prompt: not big weights needed. Wrap your text between "(masterpiece, best quality)," and the LoRa "add_detail" with a weight of 0.7 ~ 1.

Neg prompt: this one will do "3 legs, (worst quality, low quality:1.4), watermark, nsfw"