Due to the scarcity of high-quality source material of her, I minimized the number of training images. Additionally, if you find the generated images don't resemble the person well enough, please use increased Lora weights, such as 1.1~1.5