LoRA for Kathy Chow 周海媚, SD1.5 based, best with BRAv7;
v4 has some more 80s foggy feelings and more accurate on face;
v5 is sharper slightly, easier create upper body image;
all are not easy to get full body image, and may generate text at corners
LoRA was trained with Kohya, based on 200 images from internet and vintage TV drama screenshot.
Willing to receive more high res image of Kathy Chow for better LoRA.
R.I.P Kathy Chow