This LoRA is trained on Japanese gravure model Mizutani Sakura (水谷 さくら). It does a pretty good job capturing her likeness.
Recommended to use Adetailer as SDXL has trouble rendering non-closeup faces.
sakuramizutani.safetensors