v2 was trained with 1,700 images to generate pictures of cute girls, with a focus on the number of fingers and toes. Seems like it doesn't work that well
aniairy_style_v2.safetensors