This is a LightWeight/Revisited variant of my workflow: SD3 / SDXL → SDXL / SD3.
This workflow is designed to be both feature-rich and user-friendly. It includes:
A global switch for SDXL/XLL, allowing you to configure everything with a single click.
On/Off switches for key functions to optimize rendering time.
Control panels for easy management, categorized for convenience.
Color-coded elements to guide the eye and renamed nodes for better clarity.
TextBoxs for Links if needed, Wildcards List, including my own Wildcards library and Luts collection
The pipeline pathing is as follows:
First Pass Diffusion
HiResFix / SUPIR / Both
Detailers Control Panel
Hands Refiner
Vagina Refiner
Face Detailer
Refine Anything (using words for detection)
Eye Detailer
Fine Details Upscale
Post-Process Control Panel
Upscale Control Panel
Save Image
Each stage has its own On/Off switch to save rendering time.
Specific prompts for Refiners.
Face, and Eye Detailers share their own positive prompts.
(Using a few related words for the area you are working on instead of the first pass diffusion prompt can make a significant difference.)
Every step of the process includes a before-and-after comparison node at the bottom of the workflow.
An Img2Img with autoprompt based on image input / Controlnet, among other cool things;)
The default template comes with custom default settings, so you just have to download the nodes and models with ComfyUI Manager and you are ready to go
Enjoy!
Description
FAQ
Comments (1)
Many thanks for the tips.
In response - a small piece of advice to replace the rare ymc node that is not needed in this case with the one that many already have from the impact pack package (ToBasicPipe).
And I replaced 2 nodes from the segment anything package with analogs from segment anything plus.
Btw, noticed a typo in Other Useful Links (Ultralyrics)