This LoRA is used to generate this specific character, whom I call Yugo. The character is a young Asian girl, and the training focuses on her facial features and face shape, that is, her appearance. Other aspects such as hairstyle and clothing can be adjusted as needed. Currently, due to the limitations of the training images, the LoRA performs well at close and medium distances, but problems such as blurry faces may occur at long distances (sometimes img2img high-definition restoration can improve this to some extent).

Yugo

During the training of v3, the number of training images was significantly increased. Additionally, more images featuring people at medium and long distances were incorporated, along with more images of people in various poses and from different shooting angles—all of which have enhanced the corresponding generation performance.Based on my initial tests, the facial restoration effect for people at long distances is much better than that of v2(with face-fix). Meanwhile, the performance at close and medium ranges remains consistent with v2.Please have a try!Hope you enjoy it!

Files with SHA256: 956000fc1988cbbbea9dab69a35aba205376fa7928ef5fbac6381d2b1bbe4037