Open dsp6414 opened 1 year ago
As far as I know, SD-XL is not trained for inpainting task, its performance is worse than sd1.5-inpainting model, here is a discussion from diffusers:
Sanster You can use controlnet inpaint model with stable diffusion models, it works with any 1,5 model and works much better than regular inpaint 1.5 model, by using controlnet inpaint you will be able to implement outpainting like new adobe AI fill https://github.com/Mikubill/sd-webui-controlnet/discussions/1597
is there any plan to support SD-XL?