Trained on a mixture of screengrabs from the show and r34 images.

To be honest it's a bit of a tricky one to get much out of, being both trained on still images and a cartoon style which Wan doesn't seem to like much. It often gives results with very little movement and doesn't mix well with other Loras. Plus the r34 reference images came through pretty strong and oftentimes will give results showing ??????? through the clothing even if not prompted.

Still, figured I might as well share it.

Haven't really tested it with I2V but it might have a bit more application there insofar as it would presumably help keep the face consistent but I'm not 100% sure.

"turanga_leela" is the main trigger word, but I always use "her purple hair is worn in a ponytail. she has 1 eye." also.