Hello eveyone! This time I'm bringing you my most experimental LoRA yet. Took me a while to get this LoRA to a state that I'm mostly satisfied with, so I hope you'll like it as well.

This is a LoRA meant to mimic the aesthetics of dryads from GWENT: The Witcher Card Game, it was trained on card arts from that game. They are female creatures living in forests, they have have green skin, don't wear a lot of clothes, and usually have brown hair with dreadlocks, though blonde hair, green hair, and red hair dryads also exist in the game.

Depending on your checkpoint, you may have to play around with LoRA weight settings, but 0.5 is should be a good staring point, but again, feel free to experiment.

Since in all the training images, you can see the full body of the character, the LoRA will naturally try to create full body images as well, especially at higher LoRA weights. To combat this, try using the prompt "closeup", maybe even in brackets, or use ControlNet.

The prompt "convenient censoring" tries to censor ??????? by covering them with her hair, and the ?????? by clever character posing. However, it won't work 100% of the time, especially if you are using the 2nd version of the prompts I gave you. You can try adding something along the lines of "breasts, ???????, ?????, ??????" to the negative prompts to combat this, but even that's not guaranteed to work, especially if your character is standing, as it is more difficult for the AI to cover up private parts in certain poses.

Much to my surprise, the LoRA actually seems to work even with realistic checkpoints, which is nice.