The keyword is "Derpy Hooves". The training was based on Anything-v5. The dataset is a bunch of different humanized interpretations of Derpy Hooves, so I'm unsure if this will have any consistenty. The upside to this method is that you have more flexibily in dictating how she looks based on your own personal preferences. (Semi-related question: Is anyone ever gonna do an exotropia lora?)