I have used approximately 40 images to train, In order to make a realistic version you need to use image2image and make several attempts
Do not use with models that are too realistic, otherwise it looks strange,
Anime models or similar are the best
WA-06.safetensors