stable-diffusion text-to-image diffusers game

BreathOfTheWild Diffusion

This model was trained on screenshots from Botw and a few from Hyrule warrior, but because I don't actually play both of them, it was pretty hard to get decent training images. The portraits still contain resemblance of link and zelda, this can sometimes be avoided by prompting "portrait of", and the model isn't exactly easily modifiable through prompting. The landscapes are pretty good though.

To reference the art style, use the token: botw style

You can also check out the higher quality, but trickier V2 on Civitai.

Gradio

We support a Gradio Web UI to run BreathOfTheWild_Diffusion: Open In Spaces

Here are some samples.

Portraits botw5.png botw1.png

Prompt used: botw style A stunning intricate full color portrait of anad, epic character composition, wearing a white dress, brown hair, by ilya kuvshinov, alessio albi, nina masic, sharp focus, natural lighting, subsurface scattering, f2, 35mm, film grain Negative: ((disfigured)), ((bad art)), ((deformed)), ((poorly drawn)), ((extra limbs)), ((close up)), ((b&w)), weird colors, blurry Guidance: 10 Steps: 30 using DPM++ 2M Karras

I'm not a prompt wizard so you can probably get better results with some tuning.

Landscape botw3.png botw2.png botw4.png

Disclaimers

License

Responsibilities

Thanks for checking out the model!