攝影與寫實

巨微比例錯置攝影

把任何大小顛倒的想法——城市上方的巨人、頭髮裡的微型城市——拍成貼近真實的強透視照片,物理上依然說得通

更新日期:

完整提示詞

英文提示詞包含建議的範例需求。生成前請調整主體和細節。原始範例輸入及生成參數尚未完全核實。

推薦模型: GPT Image 2

Create a physically believable photograph built on an extreme inversion of scale, staging the size contrast described in the user request: a giant person towering over an ordinary city crosswalk, a miniature city hidden in someone's hair, a tiny figure dwarfed by a colossal creature or object. Sell the illusion with forced perspective and honest photographic craft: choose one coherent camera position — an extreme top-down angle or a low wide view both work — keep occlusion and contact points correct where large and small elements meet, and separate foreground, midground, and background cleanly. Light everything with consistent natural daylight and soft realistic shadows; no fantasy glow or dramatic stylization. Ground every surface in documentary detail: real skin texture with visible pores, worn streets, slightly imperfect materials, unbranded everyday objects and generic signage, bystanders at true human scale for contrast. If a reference photo is provided, preserve the person's exact face, skin, hair, and proportions without beautifying. Maximum photorealism, no AI-art sheen, no plastic skin, no extra limbs, no text, no watermark, no logos. User request: a giant woman taking a selfie above a city crosswalk, buildings only knee-high

官方範例

巨微比例錯置攝影
範例模型: Seedream 5.0 Lite
巨微比例錯置攝影
範例模型: Seedream 5.0 Lite

來源與署名

由 OmniArt 改編及重寫,以下列出來源署名、修改說明及使用條件

  • 星云巨鲤与小人的奇幻对话 · @liyue_ai

    freestylefly/awesome-gpt-image-2

    Repository code is MIT; its disclaimer says community content belongs to original authors/platforms and commercial reuse is not guaranteed.

    Rewritten and consolidated into an OmniArt recipe archetype; template tokens removed; named people, brands, platforms, games, and film IP generalized; no raw prompt text or source image copied.

  • 俯拍巨女城景自拍 · @saniaspeaks_

    freestylefly/awesome-gpt-image-2

    Repository code is MIT; its disclaimer says community content belongs to original authors/platforms and commercial reuse is not guaranteed.

    Rewritten and consolidated into an OmniArt recipe archetype; template tokens removed; named people, brands, platforms, games, and film IP generalized; no raw prompt text or source image copied.

  • 头发里的微型城市 · @krafterlab

    freestylefly/awesome-gpt-image-2

    Repository code is MIT; its disclaimer says community content belongs to original authors/platforms and commercial reuse is not guaranteed.

    Rewritten and consolidated into an OmniArt recipe archetype; template tokens removed; named people, brands, platforms, games, and film IP generalized; no raw prompt text or source image copied.

來源與署名

相關的提示詞配方