GPT Image 2: Isometric Product Scene (3D Diagram for a Product or Process)

【用途】[使用場合] 用的等角立體示意圖，主題＝[場景主題]。
【視角】嚴格等角投影（true isometric），物件以 2:1 等角網格擺放，無透視收斂、無消失點，俯視約 30 度。
【主體】用微縮模型（miniature diorama）的方式呈現 [場景主題]：把關鍵元素 [要放進場景的元素] 排列在一個方形小平台或小島上……

[ Log in to see the full prompt ]

Isometric illustration most often breaks because the AI sneaks in real perspective — the image ends up looking slightly off, with parallel lines converging when they shouldn't. So the prompt needs to strongly emphasize 'true isometric, no vanishing point, aligned to a 2:1 isometric grid' — GPT Image 2 handles this kind of geometric constraint reasonably well once it's spelled out clearly. Framing it as a 'miniature diorama island' works well because it naturally contains the elements within a bounded area and creates layering automatically, which is much more controllable than asking for a whole open scene. 'Soft ambient occlusion' is the key phrase that gives it dimensional depth without turning into a cheap-looking plastic render. The same logic applies to telling it not to add text — in isometric scenes, AI loves generating fake signage text on its own, and it's almost always wrong. This style has excellent cost-to-value for a website hero or a feature-section graphic.