It turned out pretty good, I had to include several low-quality images in the training sample so that the lora, although she roughly understood how to draw hair, fortunately this did not greatly affect the quality