For best outputs:

Version 1.5:

Sampler: DPM++ 2M SDE.

Steps: 36, cfg 7, CLIP skip 2.

VAE: kl-f8-anime2_fp16.safetensors.

Pos prompt: add the LoRa's "add_detail" (weight 0.7) and "FaceBeauty_qinglong_V3" (weight 1.0) at the end of your pos (included in my both versions' example runs).

Neg prompt: "(nsfw:1.2), (nipples:1.2), (breasts out:1.2), (nudity:1.2), NG_DeepNegative_V1_75T, (worst quality, low quality:2), (malformed hands, poorly drawn hands, mutated fingers:1.4), bad-image-v2-39000, loli, gape, bad-hands-5, drawing, sketch, illustration, CG, digital art", being the embeddings "NG_DeepNegative_V1_75T", "bad-image-v2-39000" and "bad-hands-5" embeddings included in my v1.5 example runs.

Version 1.0:

Sampler: Euler a.

Steps: 50, cfg 7, CLIP skip 2.

VAE: blessed2_fp16.safetensors.

Pos prompt: add the LoRa's "add_detail" (weight 0.8) and "FaceBeauty_qinglong_V3" (weight 1.0) at the end of your pos (included in my both versions' example runs).

Neg prompt: "EasyNegativeV2, (By bad artist -neg:0.5), (worst quality, low quality:1.4), (malformed hands:1.4), (poorly drawn hands:1.4), (mutated fingers:1.4), smile, nude, nsfw, bad-hands-5", being the "EasyNegativeV2", "By bad artist -neg" and "bad-hands-5" embeddings included in my v1.0 example runs.

What changes from v1.0 to v1.5: input_image feature, overall quality and realism increased.

Checkpoint Merge.