A lora of Tomie Kawakami. Almost all training images had the same pose, so flexibility is not that great. Also, almost all images were of either upper body shots or portraits. I literally only found 1 image with a full body shot.

So if you want to create a full body image with this lora, you need to first generate a full body image, open something like photoshop, then simply paint over another face beside her own face. Send the image to img2img, and use a denoise strength of 0.4 or 0.45. This is only for full body shots since the training data is lacking in that aspect.

Recommended settings:

Strength: 0.8-1.0

Prompt: tomie, another face, deformed face

(You can also add stuff like horror, creepy, creepy eyes etc.. but these usually change stuff i don't want. Horror tends to make zombies for some reason).