Write one image-generation prompt for a coherent environment reference: depth, boundaries, openings, landmarks, and object placement clear enough for a downstream video model.
