This LoRA is trained using some image taken from the game. All of the images used are at largest 640x480. It does not really give the satisfying results yet, but it can somewhat mimic what 90's visual novel art style.
pia3.safetensors
Same model published on other platforms. May have additional downloads or version variants.