First ever model created, honestly turned out better than I expected. Trained on over 100 images of panels from the 1st volume of the Coroika Splatoon manga. It can actually do images in color, despite the dataset being mostly in black and white, but you'll have to get creative with it. Also tends to create odd mix n' match cheap knockoff versions of the original characters...I'm going to have to fix that in a future version with more careful captioning and a bigger dataset. For now lowering the strength might give you more variable characters. Use "coroika splatoon" to trigger the lora. It does ok at making regular images, but using "comic" in prompt pushes it in the manga panel direction.