This lora is trained using Kohaku-XL gamma rev2 as the base model and has been tested to work under Animagine XL V3 (with reduced weight, around 0.8). The effectiveness when used with other models is not guaranteed.
This lora is designed to reproduce the art style of the Persona series (Soejima's style) within the SDXL anime model. I do not own any copyright for training or using this model to generate images. Please do not use this model to produce images that infringe copyrights. I will not be responsible for any copyright issues arising from images generated by this model.
The model aims to restore the art style, not necessarily to replicate character features accurately. However, you may try using: character name + character features as prompts.
For example: For Aegis: aegis_(persona), blonde_hair, blue_eyes, highres, joints, persona, persona_3, ribbon, robot, robot_joints...
Even so, it might still be necessary to combine this lora with others to accurately replicate characters.
(For character names, please use booru tags.)