Was just seeing this : https://www.linkedin.com/posts/varunkumarsiddaraju_3dmodeling-3d-ai-ugcPost-7130747356974616576-98iE?utm_source=share&utm_medium=member_desktop Looks interesting
If your level design block out is in 3D you can render out a depth map image and feed that back into stable diffusion using it's control net extension. It's possible, but the workflow is very rough and not user friendly.