






First and foremost thank you to @konyconi and this amazing LoRA guide. I was able to take the config settings and apply them to a Kohya LoRA Dreambooth collab as I need a much better GPU. The captioning tips were key I believe and I think there is still room for improvement on my end.
This LoRA was trained with 50 images of women with copious amounts of ????? on their face. In my testing it will cover the entire face, but struggles with other body parts, hence the name.
Anywhere between 0.4-1
0.4 tends to be a more conservative amount, singular load range
0.6 or more gets you into the ??????? range
When paired with other adult themed models/LoRAs you may want to bring the weight down. Works fairly well with every modal I tested.
cohf, ????? woman face covered in semen
Does a decent job of inpainting, leave restore faces off as it removes everything this LoRA adds.
more training images, would like to get to the 100+ range
I would like to cover other parts of the body, but that may require a name change
simpler captions when training, "face covered" in prompts will occasionally get you an image of someone wearing a mask
1. 轉载模型僅供學習與交流分享,其版權及最終解释權归原作者。
2. 模型原作者如需認領模型,請通過官方渠道联系海藝AI工作人員進行認證。我們致力於保護每一位創作者的權益。 點擊去認領
