very basic workflows for sdxl text to image, it has a main t2i generarion workflow, 4 util workflows---a tagger, a upscaler, a image metainfo reader, a object cleaner.

The main t2i gen workflow is modified from official impactpack face detailer workflow, with stack lora loader, tiled vae, sdxl aspect ratio selector added.

the 8 step generation is achieved by using lora "WAI-illustrious-Rectified-4Steps", if you want fast generation don't skip lora group

the eye detection model can be either download from huggingface

https://huggingface.co/ashllay/YOLO_Models/blob/main/bbox/Eyes.pt

or suggest resource section down below

update:

0.1.9

fix postprocess comparison group

0.1.8

add description to cleaner

split t2i to 2 seprate wf