怎麼填變量：[1870] 可以換成 1900、1930 等年份，年代越早畫面越模糊、越偏褐色；場景、人物、動作三處分開寫，越具體越像真照片。人物不要超過 3–4 個，人多了容易出現肢體錯亂。

常見問題：

• 太清晰、像新拍的：把"大幅降低清晰度"再強調一遍，或加"輕微重影、長曝光留下的拖影"。
• 破損蓋住了主體：把"斜向裂口"挪到畫面邊緣，例如"右下角有一道裂口"。
• 出現現代元素（電線杆、塑料製品）：在最後一句裏點名排除。

提醒：這是生成"並不存在的老照片"，適合創作和教學；發佈時請註明 AI 生成，不要冒充真實歷史影像。

英文原版

text
Non Existence Vintage Photographs with GPT Image 2 📸

- Prompt 👇
a photographic image in the style of 1870, [SCENE DESCRIPTION], with [CHARACTERS described in period accurate clothing], [Describe the interaction].

The photo has an aged and worn appearance, as it was taken in 1870. It features prominent time-induced chemical stains, heavy grain, sepia toning, and deep scratches.

Significantly reduce the sharpness so that the details of the [SUBJECT] are not crisp, making the [SUBJECT] blurry and low-fidelity.

Greatly increase the wear of the photo, including small tears, missing corners, water damage, and small wormholes caused by insect damage. Add a prominent, jagged diagonal cut across the photo, mended clumsily with old, discolored tape.

改編自 @Arminn_Ai 發佈、EvoLinkAI/awesome-gpt-image-2-API-and-Prompts 收錄的提示詞，倉庫許可證 CC0 1.0。