The basic terrain image is generated by an AI tool called NVidia Canvas, which makes it easy to paint hills and other natural features and generate photographic-quality output. It is an incomplete Beta test product, so cannot do human features such as roads and buildings, is limited in the output image size (4096 pixels, about 2 feet at 150 DPI), and can take a lot of coaxing to get a satisfactory result. However, with a bit of effort, it can achieve results that would otherwise be beyond my skills.
I intend to cut and paste buildings from other images to populate the scene, draw or paste on roads, and 3D print some low relief smaller scale buildings for a forced perspective effect. We will have to see how it turns out, and I may yet end up scrapping it and trying something else.