怎麼填變量：上載主體清晰、背景不太亂的照片，建築和機械類效果最好。[單色藍圖系統] 可以指定顏色，比如"紅色線稿系統"（示例裏東京塔就是紅色）、"墨綠色系統"；[3:4 豎版] 也可以改成"4:5 豎版"。一次上載多張照片會逐張出圖。

示例圖兩張：東京塔的黃昏照片配紅色線稿結構圖（主立面、平面圖、桁架節點、剖面小圖，標題"TOKYO SIGNAL TOWER"）；天壇祈年殿配淺紅色測繪圖（立面、斗拱、台基剖面）。

常見問題：

• 下半變成普通線稿、沒有"內部結構"：加"必須出現剖面或爆炸圖，至少 3 處結構標註"。
• 上半照片被重畫：強調"上半原樣保留，只調色"。
• 人像照片：模型會畫成"姿態 / 動線分析"，不會畫解剖，但效果不如建築和物品穩定。

適合：建築與城市主題海報、設計作品集、科普帳號配圖、展覽視覺。

原版提示詞

text
Please turn each uploaded photo into an independent high-end design poster. Use a {argument name="composition" default="3:4 vertical layout"} with two equal horizontal sections, each occupying 50% height.

The upper half preserves the original photo with subtle high-end color grading for a fine-art gallery feel. Environment can be extended naturally but the subject remains unchanged.

The lower half extracts the most recognizable subject and reconstructs it as a transparent structural blueprint and technical diagram. It reveals representative internal structures, connections, material layers, or spatial logic, creating a visual language between engineering drawings, X-rays, and futuristic archives. It interprets 'internal structure' based on the subject: components for objects, spatial construction for architecture, or motion paths for animals, avoiding fleshy anatomy.

The composition centers on the subject using transparent layers, exploded views, and fine annotations. Use fine lines and clear white space to maintain a clean reading path. The color palette is derived from the main color of the top photo, converted into a clean, professional {argument name="monochromatic system" default="monochromatic blueprint system"} (blue, green, red, etc.) with transparency and highlights for depth.

Text is transformed into technical annotations and editorial layout. Short titles and structural labels are integrated using lead lines and arrows, maintaining the restrained aesthetic of high-end magazines. The overall result is a combination of engineering order and editorial beauty, suitable for any subject from architecture to nature.

改編自 小小東 發佈、YouMind-OpenLab/awesome-gpt-image-2 收錄的提示詞，許可證 CC BY 4.0，本站已翻譯並改寫（修改內容見 changes）；示例圖來自同一條目。