It's confusing that after I set training images larger to 640*704 and remained other settings like before, I thought I would get better results, but the outputs after hirex.fix (448*640-->716*1024, denoising 0.45, latent) look not like her, sometimes on the contrary.
By the way, I happened to find that ESRGAN4x may works better than Latent or Latent (bicubic antialiased) while keeps the same high speed of these two? Could u share me with your hires.fix settings in comment? Thanks!
Weighted around 0.8 as ever.
~~
Be kind to these lovely.
Legal & Risk
It is prohibited to use this model for commercial purposes and any scenarios of illegal acts and purposes.
Description
Trained on vivid Mix, 20 epoches and 1e-5.
FAQ
Comments (6)
Thank you for this!!! Really appreciate it after requesting. I'll try the suggestions u said.
太像了，99%了，本人来了都要吓一跳，不会被投律师函吧
一天没看就这么像了大佬牛逼
为什么之前几个CHINESE模型都没了
wooo
a newbie question here, is the safetensors file supposed to be put under extensions\sd-webui-additional-networks\models\lora? my generated pictures look nothing like her :(