one image lora of character Wuwa new 4star character Youhu

The training is based on a single image, which is unavoidably a bit overfitted, but the generalization via the reg method was quite usable in my tests, or use loractl