I created this workflow for basic text2img purposes, and would like to share this to everyone who wants a simple and usable workflow without any extra stuff.To use the Groq LLM nodes, make sure to follow the instructions here: https://civarchive.com/articles/4997 (I removed the Groq nodes in the newer versions).
Check "About this version" to find the incremental changes with each version.
LATEST:
Made the workflow groups completely modular, with input and output reroutes for every group.
Complete revamp of the workflow. Made things much easier for beginners to understand.
Added support for controlnets. (I recommend controlnet union for SDXL)
Added support for simple Regional Prompting using Comfy Couple nodes.
Added different samplers for AYS, non-AYS, and SD3, and made it easier to switch between them.
Face detailer and Hand detailer nodes have tweaked values for overall better performance.
Switching between SDXL and SD1.5 LoRAs should be easier now.
Description
Tidied things up.
Added better labels and notes for beginners.
Added hand detailer
Images are adjusted with filters after final upscale to restore contrast and colors lost during the process.
Added a much better LORA loader.
FAQ
Comments (2)
I can't seem to make it work. It asks for the PerturbedAttention node, but i install it via the Manager, and it doesn't work at all
Download the extension manually from here: https://github.com/pamparamm/sd-perturbed-attention
Either clone the repo (or just download and place the folder) into "ComfyUI\custom_nodes"
Or you could also entirely remove that node, and just connect the "model" output from "Automatic CFG" directly to the "model" input of "Power LORA Loader (rgthree)" node in the LORA group.
I would recommend just removing it entirely, as it didn't work out too well for me in terms of prompt following when I used it. But some people swear by it, so I just kept it there.