Five-Panel Benchmarking Challenge
A complex multi-part prompt used to benchmark AI performance across five different styles and subjects, including clocks, humans, liquids, and games.
适合生成网页、App 或仪表盘界面稿：先保留模块结构，再替换成你的业务文案和品牌视觉。
画幅建议
16:9 或 4:3
细节策略
先快速出图，再细节精修
生成流程
先用文字生成定主画面，再一次只改一个重点细节。
完整 GPT Image 2 提示词
英文
Create an image consisting of {argument name="panel count" default="five"} sub-images, each with a different style: 1) An analogue clock showing {argument name="time" default="7:42"}. 2) A woman holding cards in her hands. 3) A glass of wine filled to the brim. 4) Real photography of a physical chessboard, fully visible, with all the pieces in their standard starting positions. 5) Two dice, with the numbers clearly visible.相关 GPT Image 2 提示词
VR Headset Exploded View Poster
Generates a high-tech exploded view diagram of a VR headset with detailed component callouts and promotional text.
Illustrated City Food Map
Generates a hand-drawn, watercolor-style tourist map featuring numbered local food specialties, landmarks, and a legend.
E-commerce Live Stream UI Mockup
Generates a realistic social media live stream interface overlaying a portrait, featuring customizable chat messages, gift popups, and a product purchase card.
Knight-Mage Battles Stone Golem
A cinematic fantasy battle scene for generating epic game-style artwork of an armored heroine fighting a giant stone golem in a ruined cathedral.
Censored Elven Adventurers in Medieval Village
Generates a cinematic fantasy scene of two detailed elven adventurers walking through a medieval village with castle backdrop and obscured faces.
Faceless Cyberpunk Infiltrator
Generates a cinematic anime scene of an armed faceless woman infiltrating a high-security futuristic server room.