Reference from Granblue Fantasy Character, Hecate.

Due to nature of SDXL Using lot of VRAM, it advise the user to adding quality prompt such as masterpiece, best quality and lot more. User may copy generate info from one of image that been posted instead.

If user still wish to utilize hires, it is advise to use with Tiled Diffusion & VAE.
For Great result, please use Animagine XL V3

Simplify about Model:

• This Model are train with Animagine XL V3 Base

• The Model are train with nsfw content. You may put nsfw into negative prompt to reduce exposure instead.

• This model lack of dataset, but it should work just fine.

• Weight only working from 1.0~0.8

• Avoid using weight below 0.7, it producing a curse gen. (If problem even came on default weight, please feedback to me)

• Recommend to use with ADetailer if Hires did not use to fix eyes detail.

How to use Lora:

• -download the LoRa Model

• -move this data into your Stable Diffusion -> Models -> Lora