首页 / 正文

上帝视角城市微缩地形图:一条提示词做出“博物馆级沙盘”质感(GPT-Image 2)

Mooko
发布于 2026-04-23 · 5分钟阅读
826 浏览
0 点赞 暴击点赞!

上帝视角沙盘城市:做出“像真的摆在桌上”的微缩地形图

你肯定见过那种巨爽的画面:

  • 整张图像一块真实沙盘
  • 山脉有起伏,峡谷有深度
  • 河流反光像真水
  • 城市挤满建筑、道路、桥梁
  • 地标高高冒出来,一眼就知道是哪
  • 侧光打过去,阴影拉得很长,质感直接封神

这类图最关键不是“城市多”,而是 “物理模型感 + 真实光线 + Tilt-Shift 景深”

下面给你一条能直接抄的提示词。你要做的事就一件:把城市名填进去。


你直接复制用的提示词(模板)

[Insert City Name] 换成你的城市名(中文或英文都行,建议英文更稳)。

A vertical, portrait-oriented, photorealistic close-up photograph of an expansive, incredibly detailed physical miniature geography and city model of [Insert City Name], filling the entire frame. The landscape is a sprawling 3D topographic map, rich with organic textures: rugged mountain ranges with diverse elevations, deep gorges, and winding scale-model rivers that reflect the light. The surface shows realistic details of earth, rock, rich foliage, and water, resembling a masterfully crafted museum diorama. Nested organically within this terrain is a highly detailed miniature metropolis. Thousands of tiny, intricate buildings, roads, bridges, and urban infrastructure are integrated into the landscape. The city's most recognizable iconic landmarks rise significantly above the surrounding architecture, serving as the focal point. Strong, natural sunlight from the side casts long, realistic shadows from the mountains and skyscrapers across the textured landscape, creating depth and drama. The depth of field is shallow (tilt-shift style), keeping the central city and its immediate surroundings in sharp focus while gently blurring the distant edges of the sprawling terrain. The entire model appears to be a tangible, physical object bathed in real light.

比例提示:你原文里写了 --ar 9:16,这更像 Midjourney 的参数写法。

  • GPT-Image 2:在生成设置里选 9:16 竖图(或最接近的竖向尺寸)。
  • Midjourney:可以把 --ar 9:16 加回去。

城市名怎么填更容易“像”

你会发现:同一个提示词,填不同城市,像不像差很多。

推荐用这两种写法:

  • 城市 + 国家/地区Tokyo, JapanShanghai, China
  • 城市 + 地标关键词(更稳):Paris with the Eiffel TowerNew York City with the Statue of Liberty and Manhattan skyline

如果你想让模型更“懂你要的地标是主角”,把这一句替换得更强硬点:

The main iconic landmark(s) are unmistakable and clearly recognizable, rising above the city as the primary focal point.

让画面更“博物馆沙盘”的关键调法

这类图最怕变成“普通航拍”。咱们要的是“桌上实体模型”。你可以在模板里加几句增强效果。

1)强化“实体物件感”

加一段:

The miniature looks hand-crafted with ultra-fine physical details, like a museum-grade architectural maquette; subtle material imperfections, tiny seams, and scale-model realism.

2)光线别用顶光,用侧光

顶光容易平。侧光才有戏。

你可以把“from the side”再写狠一点:

Golden hour side lighting, long crisp shadows, high micro-contrast on terrain textures.

3)Tilt-Shift 景深要“中间清楚、边缘虚”

想要那种“看起来像微缩玩具”的味儿,景深写清楚:

Tilt-shift miniature effect, shallow depth of field: sharp focus on the central landmark area, gradual blur toward the far edges.

可直接套用的 3 个示例(省得你改)

示例 1:东京(地标强一点)

A vertical, portrait-oriented, photorealistic close-up photograph of an expansive, incredibly detailed physical miniature geography and city model of Tokyo, Japan, filling the entire frame... The main iconic landmarks (Tokyo Tower and Tokyo Skytree) are unmistakable and clearly recognizable, rising above the city as the primary focal point... golden hour side lighting... tilt-shift miniature effect...

示例 2:上海(外滩+陆家嘴)

... miniature geography and city model of Shanghai, China ... iconic landmarks (Shanghai Tower, Oriental Pearl Tower, Lujiazui skyline) clearly recognizable ... winding river reflecting light ...

示例 3:巴黎(埃菲尔铁塔绝对主角)

... miniature geography and city model of Paris, France ... the Eiffel Tower is the dominant focal point ... museum-grade diorama ...

你想玩花一点?这几种“情绪版本”很好用 😎

把下面任意一段追加到主提示词末尾就行。

  • 雨后湿润质感

    After-rain atmosphere, subtle wet sheen on roads and rocks, soft reflected highlights.
    
  • 夜景霓虹版(更赛博)

    Nighttime diorama lighting with miniature city lights and neon glow, still photorealistic, long exposure feel.
    
  • 冬季雪景版(层次更明显)

    Winter setting, thin snow dusting on mountain ridges and rooftops, cold sunlight, crisp shadows.
    

避坑清单(翻车基本都在这)

  • 看起来像航拍:你没强调“physical miniature / museum diorama / tangible object / real light”。补上。
  • 城市太稀疏:把 Thousands of tiny, intricate buildings 保留,别删。
  • 地标不明显:加“unmistakable / clearly recognizable / primary focal point”,还不行就点名地标。
  • 画面没层次:把光线改成“side lighting / golden hour / long shadows”。
  • 全图都糊或全图都清楚:Tilt-Shift 需要“中间清楚、边缘渐虚”,别只写一句“shallow depth of field”。
  • 山河像塑料玩具:强调 realistic details of earth, rock, rich foliage, and water,再加“micro-texture / subtle imperfections”。

一句话工作流(照做就行)

选 9:16 竖图 → 粘贴模板 → 填城市名 + 点名地标 → 加侧光 + Tilt-Shift → 不像就只改两处:地标描述光线

你把城市名发我(比如“成都/Chengdu”),我也可以顺手帮你把“地标点名版提示词”写到可直接生成的程度。

OpenClaw
OpenClaw
木瓜AI支持养龙虾啦
木瓜AI龙虾专供API,限时领取免费tokens
可在 OpenClaw接入全球顶尖AI大模型
立即领取