ComfyUI Workflows - MarkLAB - CivArchive (CivitAI Archive)

Uuhh... welcome to my lab. *ehek... scratches head* I don't usually invite people down here... Actually, I don't even come down here that often myself. Mostly when I run out of buzz... or when I need to access checkpoints that aren't available on-site. Oh, also when I need to generate a trained LoRA. I built all of this myself! >.< and I always use it. It may look a little... messy. Well... go ahead. Take a look around.

MARKGEN - FLAME

aka Draft Mode. Suitable for low-VRAM hardware to test prompt generation before processing the image further.

MARKGEN ULTIMATE - BLIZZARD

SDXL workflow to generate images, upscale, and improve details

APPARATUS (what you need)

Custom nodes installed through ComfyUI manager (rgthree / mikey / was / impact pack / impact subpack)
Fdetailer for furry-type character face detection (NSFW)

EXECUTE WORKFLOW

We're not going to click the 'play' button in the top-right. Instead, click it on the node.

EVERY AFTER-PROCESSING EXPLAINED

Load Image – There are 2 image loaders (import image & select image from generation)
Latent Upscale – Upscale and create more detail in the image.
Hi-Res Fix – Upscale image using model to a high resolution and furnish into high-quality image. Stay in low denoise value since this process build unstable anatomy.
Face Fix – Does not upscale. Automatically detect and build more details to the face. Sometimes face cannot be detected at a certain head pose (rare).
Upscale to 4K – Enlarge image to 4K resolution.

Using Bypasser is not possible since it will say there are missing inputs. If it had worked, the workflow would be much easier to use by simply toggling off image processing. The node stays there just to let you know.

Instead… connect nodes manually (blue line). Pay attention to where the flow goes.