This is the workflow that I'm currently using to generate new images in SDXL for https://aipromptguide.com/

I'm keeping it here so I won't lose it, and so anyone can replicate any of the images on the site.

If you want to make your own SDXL workflow, this is a good start.

I put in all the best tech that I could, without messing with the base model's style. It uses Kohya Deep Shrink, Perturbed-Attention Guidance, and PerpNeg Guider (although the later does nothing for me, since I don't use a negative prompts for images on the site to run perpendicular to). The sample images are base SDXL with no negative prompt.

How to use (as a developer):

1. Enable developer mode in the ComfyUI settings.

2. Add --listen to your ComfyUI launcher.

3. Click Save (API Format)

4. POST the modified JSON in the body to  http://127.0.0.1:8188/promp.

There is a web socket that provides real time queue statue and progress ws://127.0.0.1:8188/ws.

GET http://127.0.0.1:8188/queue to get the entire current queue.