The LORA model used 525 character portraits from Rance X -決戦- and some game CG images for training, completing 7860 steps.

Please try to avoid using checkpoints with significant style differences.