Remember to use the trigger word seolyoona. I used LORA weight 1 and Hires fix with latent bicubic antialiased, 20 steps, 0.5 denoising strength, and 1.25x upscale. I've used ChilloutMix, VisionGen, and LOFI to generate images. Some of my example pictures use inpainting at full resolution on the face (only masked at 512x512) to improve the face quality, though you generally don't need denoising higher than 0.5.
I've trained a lot of different versions and this one is the one I like the most, trying to avoid overfitting as much as possible while still maintaining the resemblance. You still might need to use ControlNet to improve the pose and inpainting on the face to improve the likeness.
1. 재게시된 모델의 권리는 원 제작자에게 있습니다.
2. 모델 원작자가 모델을 인증받으려면 공식 채널을 통해 SeaArt.AI 직원에게 문의하세요. 저희는 모든 창작자의 권리를 보호하기 위해 노력합니다. 인증하러 이동
