I've been trying to get flux to make realistic photos with good artistic compositions and deep field of view. This lora was trained with over 150 fine art photos (many fine ???????? but not ??????? or ????) captioned with Mistral Vision which I used for free via comfyui and their lowest api key. So this is partly an experiment on how Mistral can write captions too. It's possibly going to generate NSFW images.

If the bodies are messed up, try setting the strength higher to like 1.5