This is my first attempt at a textual inversion. 5000 steps on 45 images. Seems to work fairly well. Tested on a handful of popular checkpoints. I don't really like deepfakes, but I do like mixing different faces to come up with something unique, so I'm always looking for unique faces. Riley has a prominent nose, small recessed chin, with large wide-spread and slightly bulging eyes, and this mixes well with other TI's. This TI skews a little young, so I was on the fence about sharing it, but if used responsibly, it can do what I intended, which is to mix things up. It can take a few iterations to come up with something decent, but overall it's been working for me. Let me know if there's issues, I'm always looking to learn and improve.