Trained on 29 images using kohya's dreambooth trainer. Sweetspot for this model is to put it at .65 this keeps the style or aesthetic of the model you are using while keeping the characters features You may also have to describe some features like the facial mark and the yellow eyes. If the horns on the mask don't appear make sure to add horns to the prompt. Make sure that clipskip is on 2 and the vae is anything v3 or what ever one you like For different poses use controlnet and openpose editor. This eliminates some the poses that this lora will fall into |