Ann Takamaki from Persona 5. Dataset was rather varied so I suspect the model will not be very competent at accurately recreating her metaverse outfit. Versions for all 5 training epochs are provided here. Higher training epochs may better recreate the character but may begin to overfit. 0.75 is generally a good starting point for the network weight. Example images for each epoch done on the same seed and settings without cherry-picking, so there may be some errors. Example images here generated using a slightly modified Corneo's 7th Heaven Mix | Stable Diffusion Checkpoint | Civitai and Corneo's Shinra26 Mix | Stable Diffusion Checkpoint | Civitai. All of my LoRAs are currently available on my Hugging Face repository: Doctor-Shotgun/ds-LoRA · Hugging Face. I recommend also adding Persona / Catherine (Soejima Shigenori) Style LoRA | Stable Diffusion LORA | Civitai if you want to more accurately re-create the game's art style. |