Kitchen Sink set up (includes everything including the kitchen sink so to speak)...

Primary non-bypassed node in default here is flux gguf with a turbo lora to speed it up a bit.  On an 8 gb vram machine with this lora portion of the workflow going I generate about 1 image a minute and 5 seconds or if I set latent batch to 2 about 2 images per 94 seconds.

you can left click and set different groups to set to always or back to bypass to turn them off.  This hefty workflow is sort of an instructional/history type thing that is studying different types of ways of creating images in comfy ui.  Flux with the Lora is fastest/best for realism now...

but we have in here Stable Diffusion 3.5 Medium, Hunyuan, Pixart Sigma, Aura Flow, SDXL, SDXL with a Refiner, and of course Flux.  

The 'regular' flux is right above the lora version.

There's a tagger to re-evaluate that one using what the ai sees in the image generated. I may try to tagger up the lora version one of these days. WD14 Tagger is a tricky node since it does weird stuff like making almost all animals un to pokemon.

Then we have a way to take the output image from all of these and hook it up to create stable diffusion video (SVD) or Moochi Video... and last but not least a nice little old school sdxl outpainting setup that uses foocus type technology in comfy for better /faster outpainting than you normally see in comfy by itself.  I have been messing with other outpaint technology that's newer but find this one compatible for low vram systems like I currently have.  Have fun.