Digitous
commented
2 years ago imho;
K-diffusion would ideally be added to txt2imghd or decouple which sampler is used from it and rely on the hlky webui instead, which adds options for all K-samplers and seems to be updating at an insane pace. it is also based on the waifu fork and adds support for embeddings etc and can be installed on top of waifu.
https://github.com/hlky/stable-diffusion-webui
Either way I support integrating it with a webui, totally behind this 100%
Since the old txt2img_gradio.py and txt2img.py are basically the same thing in the waifu fork of stable diffusion it might be worth integrating the txt2imghd changes into the txt2img_gradio.py script for the web interface.
I did a quick diff and tried to copy over the new the txt2imghd additions, but it looks some refactoring has to be done to make txt2imghd work with gradio. Luckily It doesn't seem like it would require dramatic changes, but it would take more time than I have at the moment to really understand the new code introduced. Just a thought!