This LoRA is modeled after how she looked in the 2010s.
"Lilly" was used as the training word, but it is not required.
LillyRoma_rank8_bf16-step04000.safetensors