Initial.

The LoRA was trained on 100 images, consisting of:

• 70 images: Event CGs accessible through the in-game gallery.

• 28 images: Character sprites, background images, and event CGs unavailable through the gallery.

• 2 images: Concept art and soundtrack artwork.

No natural-language captioning was performed. Captions consist almost entirely of Danbooru tags.

The LoRA has learned the general art style and some basic character appearance features. However, it has not fully captured the game's grotesque atmosphere, and reproducing the infamous food served in a plastic container remains difficult.

dim32/alpha32, CAME, UNet LR 3e-05, cosine 900 steps, 100 images (1024²)