Using M.H. to control a text-to-image AI

[green_tomato]

It's possible to use MakeHuman to control Stable Diffusion using ControlNet's "depth" mode by rendering out the depth buffer (or "Z buffer" as the depth information is often called) using Blender or a similar tool.
This will produce an image where objects that are close will be white, and areas that are far away are dark.
Then you can use the AI's usual tools to control the style. Here are some examples:

[joepal]

That's actually pretty interesting.

I hadn't thought of using MH as input data for AI.

[joepal]

Approximation of the same idea in blender: Create a "camera distance material"

[green_tomato]

Oh neat trick. I just used the compositor section to output the "Depth" layer, like this: