ReignSoles V3.0 (Anima Base 1.0):
Finally, version 3.0 is out. In this version, I massively up sized the dataset (from ~172 images to 417). Around half of the dataset was tagged with a combination of natural language and danbooru style tags, and the other half was just tags. It is now a LoKR now as well, a new advanced type of LoRA that boasts higher performance at a much smaller size. It's compatible with ComfyUI, and it should be compatible with other supporting UI's like SD Web UI Forge Neo.

Similar to prior versions, the tags "perfect feet" and "detailed soles" were added to each image with the highest standard of quality I was looking for in both foot shape (perfect feet) and detail (detailed soles). This model was trained in a way where images were resized to fit into a common image size / resolution, with the most common resolution by far being 832x1216. I recommend using this resolution alongside the model for the best results.