Summertime Saga style lora trained for 2800 steps on about 130 images. There is a bias for characters to look to the side due to the nature of the game but this can be prompted away. Not totally satisfied with results but I figured I'd post anyway.
Summertime Saga_v1.safetensors