So I have been trying all kinds of prompts with GPT Image 2.0 to see how much I can push it. It can technically do 20x20 but it struggles with the text. It does well with 15x15. Here is a 16x16 image: 

Create a 16×16 grid with  equal cells, each showing a different culture, civilization, or region-specific scene from a different time period in human history. Each cell should contain a miniature diorama featuring distinct clothing, architecture, tools, food, symbols, and environment. No culture should repeat in the same form, and no two cells should look compositionally similar. The image should feel like a dense visual atlas of humanity with accurate regional and historical differentiation.

来源许可/权利说明：聚合仓库上游许可（请按各条目来源遵循）