Sharing the workflow I was using on ComfyUI
I started with Automatic1111, then moved to SwarmUI, and switched to ComfyUI a few months ago. I've been using comfy-based UIs to generate images for about a year, but recently I switched completely to reForge, which I used only for inpaint.
I've never really put much effort into my workflow; from using a k-sampler, it simply became more organized and included additional features like ControlNet and Upscale as I learned them.
I've never used heavy nodes, partly because my hardware can barely handle them: it struggles even with that 1.5 upscale, but due to a problem with the node itself, I had to set it to 2, which further degraded performance.
I've noticed I don't have this problem with reForge, which is why I'm switching to that UI.
So I'm sharing my workflow here on CivitAI to avoid losing it, since I don't think I'll use comfy-based UI anymore until I have hardware that can handle all the power that comfy and similar can offer.
What it does?
Nothing special.
It has a ControlNet and Upscale system.
I set up some group control nodes to quickly enable/disable the preprocessors, and another to switch from generation to upscale.
Since I've always edited the images, removing artifacts and some errors manually before upscaling the image, it doesn't make sense to me to directly upscale a raw gen.
I have inserted some purple text nodes into the workflow, which help understanding.