Use it just like any other LoRA, trained with AI-Toolkit, 30 images in 1024², 4000 steps.