If you want more detailed control and design, you need to activate the text nodes I bypassed below and make the necessary connections.
Description
FAQ
Comments (2)
I don't understand how you got a square image in the output? I get a horizontal wide stretched image. It's not surprising, the input to the sampler is two blended images. Why in your screenshot, the image is square with people in full height? How?
you can set size with an EmptySD3LatentImage