MiDaS uses trained AI to take a normal 2D image and output a 3D depth-map. In Poser-speak it’s like Poser’s ‘auxiliary Z-depth’ pass or render.
Free and public, no sign-up needed. Just drag-and-drop your image. It can probably also be installed locally, though I haven’t looked at the requirements for that.
Once you have it you can use the usual Photoshop layer inversion/blending-mode tricks to create ‘depth-fog’ in the scene, where there was none before.