4-Area Composition

Here are examples demonstrating the ConditioningSetArea node. You can load these images into ComfyUI to obtain the complete workflow.

This image contains 4 different areas: night, evening, day, morning.

Zone composition with Anything-V3 + second pass with AbyssOrangeMix2_hard

This is what the workflow looks like in ComfyUI:

This image contains the same areas as the previous one but in reverse order.

Adding a subject to the bottom center of the image by adding another area prompt.

Increasing Image Consistency with Zone Composition

Stable diffusion creates its most consistent images when generating square images with resolutions close to 512x512. But what if we want to generate an image with a 16:9 aspect ratio? Let's generate a 16:9 image with a seated subject. If generated normally, the success rate will be low, with limbs stretching unnaturally across the image and other consistency issues.

By using Zone Composition with a square zone for the subject, the consistency will be higher. Since it is generated simultaneously with the rest of the image, the overall image consistency will be excellent.

This workflow uses Anything-V3, and it is a two-pass workflow with the zone composition used for the subject in the first pass on the left side of the image. The reason for the second pass is solely to increase the resolution. If you are satisfied with an image of 1280x704, you can skip the second pass.

Adding a subject with red hair with a zone prompt on the right side of the image.

First pass output (1280x704):

Second pass output (1920x1088):

This second pass output image illustrates one of the behaviors of Stable Diffusion. The second pass has no zone prompts. You’ll notice that the hair of Subject 1 is blonde with pink highlights, and Subject 2 has pink hair instead of red hair as shown in the first pass output. This is because Stable Diffusion tries to make the overall image consistent with itself, and one of the side effects is the merging of hair colors.