Benchmark Prompt for Nano Banana/Gemini Image Generation
A futuristic educational visual explaining five key quality checks for generated images, focusing on text accuracy, anatomy, material realism, lighting consistency, and prompt adherence. The scene likely features an elderly person and an android with mixed materials and complex reflections, presented as a detailed test case.
Model: Nano Banana ProCategory: Infographic/Edu VisualStyle: Cyberpunk/Sci-FiLanguage: ja
Prompt
Five Challenges to Check with This Prompt If this single image is generated, zoom in and check the following points. - Text consistency (AI EVOLUTION 2026) Is the text written on the cardboard sign rendered correctly, without spelling mistakes, while still preserving a handwritten-like texture? - Anatomical accuracy and interaction Are the old man’s “10 fingers” naturally holding the fragile glass butterfly? Are any fingers blending into the background or the count incorrect? - Differentiation of textures (material mix) Are materials with completely different reflectance and transmittance properties—“the old man’s skin (wrinkles),” “chrome plating (mirror-like),” “glass (transparent),” “velvet (fuzzy),” and “liquid (steam)”—depicted consistently under the same light source? - Spatial and light calculation (ray-tracing-like understanding) Is the android’s blue glow correctly reflected in the old man’s glasses and nearby metal? Are the shadows cast by multiple light sources consistent? - Prompt adherence and attribute separation Is the layout “old man on the left, android on the right” being followed? Is color bleeding occurring, such as the “red velvet” color transferring to the android’s skin or the old man’s clothing turning blue?