Kofi: https://ko-fi.com/ericzhu

This version is trained on Realistic Vision, so normally would work better in real person images. Please check out showcase images to see which one works better for you.

If you want to work with more anime style, please go check out previous version of this model.

Use ControlNet would make life a lot easier, but the reference image will greatly impact the output image, causing prompts to not take effect in some cases.

Don't set sampling steps too high if you don't want the the texture of human skin to look like a pile of slime.

Please like the model and show us your creations below.

A kind review would be really appreciated.