LoRA is trained on anime screencaps. Always use ADetailer to get detailed eyes. Recommended lora weight:1.
Akane_Sonozaki.safetensors