Tint, is a full chroma finetune on 30k transformation related images, things turning into other things. It can do sequences, comics and single images around the subject.
Tint comes with several associated loras and a structure control model.
For tagging, it uses images with either llm tags or my own prompts that are basically just two-five sentences saying "transformation sequence x into y" "transformation comic x into y" "before and after x into y" or just a simple literal description for single images
You can use first/last frame references and structure control with the models in the kofi page: