最近在 X 上又有一个很火的提示词——打造超真实感美女摄影照片。

看到这个，我立刻想到之前我写过的那篇 豆包生成真实美女生活照自拍，

这一版明显是它的升级版。不同的是，这次生成的照片真实感更强烈，几乎可以乱真。

我自己也拿 Gemini 2.5 的 Nano banana 模型试了一下，确实惊艳。

画面不仅在光影质感上更接近真实摄影，还能模拟社交媒体（如抖音、小红书）常见的审美风格。

下面来看看这位博主的效果和我测试时整理出的几种版本。

真实感强，放到社交媒体平台上几乎无法分辨是AI生成的，尤其适合做瑜伽、健身类生活照风格。

还有升级版：

原文中的JSON提示词：

核心提示词结构（JSON 版）

使用的是 Google Gemini 2.5 Nano banana 模型，提示词以结构化 JSON 形式呈现。

JSON 提示词的好处是结构清晰、参数细致、复现性高，特别适合反复微调和版本对比。

{
  "style": "High-key studio portrait, direct flash aesthetic, East Asian social media style (e.g., Ulzzang, Douyin), stylized beauty retouching.",
  "output": {
    "color_profile": "sRGB",
    "render_intent": "photo"
  },
  "subject": {
    "category": "human",
    "gender_presentation": "female",
    "ethnicity": "East Asian (e.g., Korean, Chinese)",
    "age_bracket": "young_adult",
    "body": {
      "build": "slim",
      "proportions": "natural human anatomy",
      "posture": "relaxed on sofa, seated casually",
      "pose": "seated, legs crossed and tucked close to body",
      "gesture": "Right hand raised, fingers loosely curled, back of fingers/knuckles gently supporting the chin and lower cheek.",
      "head_tilt_deg": 5
    },
    "face": {
      "expression": "Playful, alluring",
      "gaze": "right eye direct to camera",
      "eye_action": "winking with the left eye",
      "skin_tone": "Very pale porcelain (lightened aesthetic)",
      "makeup": "Stylized K-Beauty/Douyin look: flawless matte base, strong pink blush high on cheeks, pink gradient lips, defined brows, light eyeliner, emphasized Aegyo-sal",
      "features": "small beauty mark/mole under the left eye"
    },
    "hair": {
      "length": "long",
      "style": "messy high updo/bun with loose strands and curtain bangs",
      "color": "dark brown"
    },
    "wardrobe": {
      "top": "white fitted cropped camisole",
      "outerwear": "light gray zip hoodie, worn open and slightly slipping off both shoulders",
      "bottom": "white lounge shorts with drawstring",
      "footwear": "barefoot"
    }
  },
  "environment": {
    "location": "studio or minimalist interior",
    "set": "black leather sofa against a plain white or light gray wall",
    "props": "Silver laptop (Apple MacBook, logo visible) placed on the cushion to the subject's right (camera left)"
  },
  "lighting": {
    "key": {
      "source": "strobe/flash",
      "modifier": "Bare reflector or direct flash (hard source)",
      "position": "Near camera axis, slightly camera-right and above eye line",
      "effect": "Crisp, dark, well-defined cast shadows on the wall directly behind subject; strong specular highlights on skin and sofa leather."
    },
    "fill": {
      "type": "minimal/none"
    },
    "ambient": "suppressed",
    "white_balance_K": 5800
  },
  "camera": {
    "system": "Digital Camera",
    "sensor": "full-frame equivalent",
    "lens": {
      "type": "prime",
      "focal_length_mm": 50
    },
    "exposure": {
      "iso": 100,
      "aperture_f": 4.0,
      "metering": "Bright exposure, high-key aesthetic"
    },
    "focus": {
      "target": "near eye (right eye)",
      "depth_of_field": "moderate"
    },
    "framing": {
      "orientation": "vertical",
      "crop": "mid-thigh to head with room above hair",
      "angle": "eye-level",
      "composition": "subject centrally framed"
    }
  },
  "color_grade": {
    "look": "Bright, clean, slightly cool tone",
    "contrast": "High contrast",
    "saturation": "moderate, emphasized pinks"
  },
  "postprocess": {
    "noise_reduction": "high",
    "texture": "Highly smoothed skin, poreless appearance ('porcelain doll' or 'beauty filter' effect)",
    "sharpen": "selective on eyes/lashes",
    "blemish_control": "Complete removal of all blemishes and texture."
  },
  "quality_targets": [
    "accurate limb lengths and joint angles",
    "correct finger count and articulation",
    "realistic fabric tension and folds",
    "accurate winking expression"
  ],
  "negative_prompt": [
    "no altered or exaggerated body proportions",
    "no extra or fused fingers",
    "no realistic skin texture, pores, or blemishes",
    "no text or watermarks (excluding specified logos)",
    "no extreme wide-angle distortion",
    "no NSFW content",
    "no dark/moody lighting",
    "no warm tones"
  ]
}

文字版提示词

如果不想使用结构化格式，也可以直接使用文字描述：

High-key studio portrait, direct flash aesthetic, East Asian social media style (e.g., Ulzzang, Douyin), stylized beauty retouching. A young adult East Asian (e.g., Korean, Chinese) female with a slim build and natural human anatomy, standing in a relaxed posture with weight slightly shifted to her right hip, legs crossed at ankles in a casually alluring stance. Her head is tilted 5 degrees, right hand raised with fingers loosely curled, back of fingers/knuckles gently supporting the chin and lower cheek in a coquettish gesture. Facial expression is coquettish and charming, with right eye gazing directly at camera and left eye winking playfully. Skin tone is very pale porcelain (lightened aesthetic) with stylized K-Beauty/Douyin makeup: flawless matte base, strong pink blush high on cheeks, pink gradient lips, defined brows, light eyeliner, emphasized Aegyo-sal, and a small beauty mark/mole under the left eye. Hair is long dark brown styled in a messy high updo/bun with loose strands and curtain bangs. Wardrobe consists of a white fitted cropped camisole, light gray zip hoodie worn open and slightly slipping off both shoulders, white lounge shorts with drawstring, and barefoot. Environment is a minimalist interior with plain white or light gray wall and light-colored flooring. Silver laptop (Apple MacBook, logo visible) placed on a small white side table to the subject's right (camera left). Lighting: key light is a strobe/flash with bare reflector (hard source), positioned near camera axis, slightly camera-right and above eye line, creating crisp, dark, well-defined cast shadows on the wall directly behind; minimal fill, suppressed ambient, 5800K white balance. Digital camera with full-frame sensor, 50mm prime lens, ISO 100, f/4.0 aperture, bright high-key exposure metering. Focus on near eye (right eye) with moderate depth of field. Vertical orientation, composition framing from head to mid-calf with room above hair, eye-level angle, subject centrally framed. Color grade: bright, clean, slightly cool tone, high contrast, moderate saturation with emphasized pinks. Postprocessing: high noise reduction, highly smoothed poreless skin ('porcelain doll' effect), selective sharpening on eyes/lashes, complete blemish removal. Ensure accurate limb lengths, correct finger count, realistic fabric tension, and authentic coquettish expression. No altered proportions, extra fingers, skin texture/blemishes, text/watermarks (except MacBook logo), wide-angle distortion, NSFW content, dark lighting, or warm tones.

更多版本

除了博主分享的版本之外，我还找到了更多的版本，一并整理并测试了

中文版

一张高调工作室肖像照，采用直射闪光灯美学，具有东亚社交媒体风格（如Ulzzang、抖音），并经过风格化的美颜修饰。
色彩配置：sRGB
渲染意图：照片
主体：
类别：人类
性别呈现：女性
种族：东亚（如韩国、中国）
年龄段：青年
身材：苗条，自然人体比例
姿势：慵懒地坐在沙发上，双腿交叉并收拢靠近身体
手势：右手抬起，手指松散地卷曲，手背/指关节轻轻托住下巴和下脸颊。
头部倾斜：5度
脸部：
表情：俏皮、迷人
凝视：右眼直视镜头
眼睛动作：左眼眨眼
肤色：非常苍白的瓷器（美白效果）
妆容：风格化的K-Beauty/抖音妆容：无瑕哑光底妆，颧骨处涂抹浓郁的粉色腮红，粉色渐变唇妆，眉毛清晰，眼线清淡，卧蚕突出。
特征：左眼下方有小颗美人痣/痣。
发型：
长度：长发
风格：凌乱的高发髻/丸子头，带有一些散落的发丝和法式刘海。
颜色：深棕色。
服饰：
上衣：白色修身短款吊带背心
外套：浅灰色拉链连帽衫，敞开穿着，略微滑落双肩。
下装：白色抽绳家居短裤
鞋子：赤脚。
环境：
地点：工作室或极简主义室内
布景：黑色皮革沙发靠着一面纯白或浅灰色的墙壁
道具：银色笔记本电脑（苹果MacBook，logo可见）放置在主体右侧（相机左侧）的垫子上。
灯光：
主光：
光源：频闪灯/闪光灯
柔光罩：裸露的反光罩或直射闪光灯（硬光源）
位置：靠近相机轴线，略微偏向相机右侧并高于视线。
效果：主体正后方的墙壁上投下清晰、深暗、明确的阴影；皮肤和沙发皮革上有强烈的镜面高光。
补光：极少/无
环境光：抑制
白平衡：5800K。
相机：
系统：数码相机
传感器：全画幅等效
镜头：
类型：定焦
焦距：50毫米
曝光：
ISO：100
光圈：F4.0
测光：明亮曝光，高调美学。
对焦：
目标：靠近眼睛（右眼）
景深：适中。
构图：
方向：垂直
裁剪：从大腿中部到头部，头发上方留有空间。
角度：平视
构图：主体居中。
色彩分级：
色调：明亮、干净、略带冷色调
对比度：高对比度
饱和度：适中，强调粉色。
后期处理：
降噪：高
纹理：皮肤高度平滑，无毛孔（“瓷娃娃”或“美颜滤镜”效果）
锐化：选择性地对眼睛/睫毛进行锐化
瑕疵控制：完全去除所有瑕疵和纹理。
质量目标：
准确的肢体长度和关节角度
正确的手指数量和关节活动
逼真的织物张力和褶皱
准确的眨眼表情。
负面提示：
无改变或夸张的身体比例
无额外或融合的手指
无逼真的皮肤纹理、毛孔或瑕疵
无文字或水印（除非指定logo）
无极端广角畸变
无不适宜内容
无黑暗/阴郁的灯光
无暖色调。

测试效果：

简洁版

画一幅东亚年轻女性的高调工作室肖像照。她穿着白色吊带背心、白色短裤和敞开的灰色连帽衫，随意地坐在黑色皮沙发上。她的深棕色长发扎成凌乱的高丸子头，留着法式刘海。她的右眼直视镜头，左眼俏皮地眨着。右手托着下巴，指尖轻触脸颊，脸上是俏皮又迷人的表情。她的肤色非常白皙，妆容是流行的K-Beauty/抖音风格，脸颊有明显的粉色腮红，嘴唇是粉色渐变，眉毛清晰，眼线清淡，卧蚕突出，左眼下方还有一颗小痣。
​
画面采用直射闪光灯照明，营造出高调明亮、高对比度的效果，在白色墙壁上投下清晰的阴影，皮肤和沙发上有明显的高光。旁边放着一台银色MacBook笔记本电脑。照片经过高度后期处理，皮肤光滑无瑕，没有毛孔或瑕疵，眼睛和睫毛部分锐化，整体色调明亮、干净，略带冷色，粉色调突出。构图是垂直的，主体居中，从大腿中部到头部。

测试效果：

扩展版

在户外草地边、阴天自然光下，一位25岁左右的东亚女性坐着，被抓拍到三分之四侧脸。她黑色马尾、空气刘海，几缕碎发贴在脸颊，有自然淡妆（轻腮红、柔和眼线、粉色唇）。她佩戴细银项链，肩颈与锁骨线条清晰，D罩杯饱满胸型，身型紧致无赘肉。上身穿海军蓝细波点吊带短上衣，心形/微V领口（深v）、金属肩带调节环可见；下身为高腰军绿色工装裤。她右手持白色毛巾，身体微前倾、目光看向右侧，表情温柔害羞、面色轻微绯红，轻咬下唇。背景为轻微雾化的草地与野餐物件（蓝色垫、橙色物件）并强烈散景，右侧有一位朋友被半截裁切且虚化。色调柔和、对比偏低，长焦镜头、浅景深（bokeh）、细微颗粒与轻雾感，胶片感但保持写实。Portra-style film grain, subtle warm tint, gentle halation

测试效果：

还有更多未开发出来的，需要自己去尝试的效果，可以看看

豆包也行

结语

从豆包的自拍照，到 Gemini 2.5 的 Nano banana 提示词，可以看到 AI 图像生成正在走向一个更精细、更专业的阶段。

JSON 结构的提示词其实就像“摄影师脚本”，你不仅定义人物，还定义光线、镜头、甚至化妆细节。