ComfyUI workflow to generate good quality images with base NoobAI vpred model and a refiner.
For convenience I moved guide for this workflow into separate article How to generate good images with NoobAI vpred.
Description
Apply sigmas from base model. Should reduce unwanted details during refiner stage.
Added prompt tokens counter.
FAQ
Comments (5)
thanks for the workflow, works very well.
i was having problems with the ApplyFBCacheOnModel but i fix it installing comfy-wavespeed
I am glad that you solved the problem. For some reason wavespeed is not properly registered in the Comfy Manager, I hope that they will fix it.
@somedoby where will be a good place to put a lora loader in the workflow?, also you know if posible to put a controlnet on the upscale to bump up the denoise ?
@rerolls26 You can apply Loras in the prompt area like a1111, it's the easiest way, Check the linked article for more info. Or you can insert lora loader just after checkpoint loader, probably also do this for the refiner model for the maximum effect.
Do you want to apply tiled controlnet? It is definitely possible, just connect it with positive and negative conditions to the upscale ksampler as usual.
You should probably disable (bypass) deep shrink node in the case for less interference.
@somedoby thanks for the reply, noted.