[PonyXL] Dakini ダキニ | Kamikatsu | Kaminaki Sekai no Kamisama Katsudou 神無き世界のカミサマ活動

Dakini from Kaminaki Sekai no Kamisama Katsudou (神無き世界のカミサマ活動)

This model was trained entirely on anime screencaps.

Base Model: Pony Diffusion V6 XL (/model/257749?modelVersionId=290640)

Prompts:-

• Trigger Words:

dakini_olis, 1girl, very long hair

• Outfit 1:

blonde hair, twintails, hair ribbon, gradient eyes, heart-shaped pupils, chain necklace, black robe, sideless dress, underboob, pelvic curtain

Optional:
ankle strap, black footwear
(^not very accurate)

• Outfit 2:

blonde hair, twintails, multicolored horns, purple eyes, symbol-shaped pupils, naked coat, chest tattoo, bodypaint

Optional:
high heel boots

• Demon Form:

demon girl, multicolored hair, pink skin, black sclera, yellow eyes, facial mark, heart tattoo, pubic tattoo, fox tail, shawl

Optional:
thigh boots

• Extra:

solo / tentacles /
anime coloring (can slightly alter the style, but may not have a significant effect)
(you may refer to the samples)

• Negative Prompt:

blurry background
(There are others, but they may not be required as long as there is no weird background bleeding.)

Known Issues:

• Minor details, especially her eyes, chain necklace and chest tattoo, are not very accurate. Her eyes in Outfit 1 and Demon Form can be improved with inpainting, but those in Outfit 2 are still not accurate even after inpainting.

• Her ankle strap in Outfit 1 doesn't look good enough.

• You probably need to add tan if you want her to be naked with tanned skin. In fact, I didn't tag her Outfit 1 with tan in the dataset because I was too blind to notice she had different skin colours in the outfits.

Weight: 1 (feel free to lower it, especially if you don't care about her outfit)

Afterword:
This was harder to train than I thought. The main reason is that I wanted to improve her eyes, so I included many portrait images with good eye details in the dataset. I even photoshopped some images to enhance the eye details. However, the eye symbol in her Outfit 2 was too hard to learn with limited images, or it might just be my skill issue.

Someone decided to update WebUI Forge at https://github.com/Panchovix/stable-diffusion-webui-reForge. I am using it now.

And...I just noticed that there are two PonyXL models of her now. There were none when I started working on her.

Hopefully some of you will like this model.