nano-banana Worldbuilding Scene Concept Art

Generate a worldbuilding scene concept image using the nano-banana (Google·text-to-image) author route, to verify the model's scene construction and atmosphere expression ability under complex natural-language descriptions. The work breaks "visual exploration of worldbuilding and atmosphere" into reusable text layers: a single English prompt simultaneously constrains spatial structure (megastructures, multi-level transit, ground reflections), environmental storytelling (fog, steam, light accents), camera and lighting (low-angle wide shot, backlit haze), and overall color palette (cool teal-gray with warm orange accents), so that a single generation yields a scene image with narrative density rather than a close-up of a single subject. Applicable scenarios include: early concept exploration for games and film/TV, visualizing location settings in scripts or novels, first-frame material for art style reference boards, and creative discussions that need to rapidly iterate scene directions in natural language. Because the author route exposes only three fields—prompt, aspect ratio, and output format—this work also serves as a control baseline: keeping the composition description fixed while swapping aspect_ratio (e.g., 16:9 for widescreen concept, 9:16 for vertical storyboards) and output_format, to observe how stable the same prompt is across different canvases. Note: nano-banana's official capability details and version naming are not consistent across public materials; this work uses only the parameters actually declared in the author schema, and makes no guarantees about undeclared capabilities such as resolution limits or editing rounds.