SDXL在lora风格的训练上有着极为出色的表现,这个瑞克与莫蒂的风格泛化性很强,基本上能与任何图像不违和的融合到一起。
测试下来,模型的权重在0.7/0.8的时候效果都不错,但是在0.9以上的话就会出现一点问题,可能是训练的步数太高导致的。
使用图生图也能有不错的效果出现。
使用sdxl的controlnet也有不错的效果。
期待大家的返图和评论,坚持每天上传一个不同风格的模型。
The SDXL has done an excellent job with lora-style training, and the Rick and Mottus styles are so versatile that they blend seamlessly with almost any image.
The test, the model's weight in 0.7/0.8 when the effect is good, but above 0.9, then there will be a little problem, may be the number of training steps is too high.
The use of graphics can also have a good effect.
Controlnet using SDXL also works well.
I look forward to your comments and comments, and insist on uploading a different style of model every day.
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
