What is the model resolution?
#3
by KKSlider2022 - opened
I'm trying to generate 1024x1024 images with this model and I keep getting four quadrants of duplicated people, implying the model resolution is perhaps 512x512. Any suggestions on how to get a single figure, just one person, centered in the image, without chimeras being duplicated and blended in to each other?
Model is trained at 512. You'll want to use something called "hires fixing" or "hires correction" to help reduce the number of duplicates you see. Both Auto1111 and InvokeAI have a version of this feature.
KKSlider2022 changed discussion status to closed