Download
1 variant available

170
1.1K
This is my simple fucking workflow (i.e., SFWF) for Krea2 in ComfyUI.

It only uses 2 custom nodes (rgthree's seed and power lora loader, because they significantly improve the user experience IMO! But you could simply remove/bypass these if you want a stock Comfy build)
This workflow is intended for txt2Img, which includes lora, and optional prompt enhancement.
The "Control Center" on the left side is where you:
Write your prompt
Choose your model
Select your output image resolution
Toggle on/off prompt enhancement
Set your seed
The right side is for everything lora:
Easily toggle loras on/off
Set the strength for each lora
Click the "i" icon for a quick peek at the Civit model page for each lora
Want to change the sampler, scheduler, steps, clip, or vae? Just open the subgraph on the "Control Center" and go wild! (these are just the settings I've found worked best for the models I use).

Why post a workflow? After my own experiences using a few of the popular "simple" workflows out there, I was genuinely curious if all those node packs and long render times were actually producing better results. For my purposes, the answer was no.
Depending on your hardware and chosen model, I'm getting ~12 seconds per image. Perfect for fast iterations through seeds, loras and prompts.
Good luck, have fun!
