GPT Image 2 人像提示词实战指南从 35mm 胶片到身份一致性的 300 案例拆解基于 awesome-gpt-image-2-API-and-Prompts【免费下载链接】awesome-gpt-image-2-API-and-PromptsGPT-Image-2 API and Prompts项目地址: https://gitcode.com/gh_mirrors/aw/awesome-gpt-image-2-API-and-Prompts本指南以仓库多语言案例库中的俄语版人像文档 cases/portrait_ru.md 为绝对主体系统拆解其中 300 个可复用人像提示词案例从 35mm 胶片街拍、霓虹夜景、CCD 闪光到 3x3 身份一致性网格、参考图驱动的 UGC 工作流、杂志版式与跨风格迁移。读完本文你将掌握 GPT Image 2 人像提示词的结构化写作方法、可复制粘贴的完整模板以及如何通过仓库的 API 示例把任意提示词直接跑成图片。案例库定位portrait_ru.md 在仓库中的角色cases/portrait_ru.md是该仓库人像与摄影案例Portrait Photography Cases分类的俄语本地化版本与之对应的英文源文件是 cases/portrait.md同目录下还提供德语、西班牙语、法语、日语、韩语、葡萄牙语、俄语、土耳其语、简体中文、繁体中文共 11 个语言版本。多语言文件之间通过仓库根目录的 script/sync_multilingual_readmes.py 维护同步关系——该脚本以README.md为源、按LOCALES字典将更新分发到各语言 README并在默认模式下保护人工维护的 News 与 Menu 段落案例文档的本地化遵循同样的结构约定。在整个仓库的七大类提示词体系人像 portrait、海报 poster、UI、电商 e-commerce、广告创意 ad-creative、角色设计 character、对比案例 comparison中人像分类是案例最密集、风格跨度最大的一类。英文 README.md 的 Portrait Photography Cases 小节会摘选部分精选案例并链接到完整文档而portrait_ru.md则保留了全部 300 案例的完整提示词原文是研究 GPT Image 2 人像生成能力的第一手素材。需要特别说明该文档定位为社区提示词案例合集每个案例都标注了原始发布来源X/Twitter 链接与作者本质是开箱即用的提示词参考而非教程式说明文档。因此本文的增量价值在于把散落在 300 案例中的共性结构、高频手法与可复用模板提炼成方法论同时完整保留关键案例的提示词原文确保信息密度不低于原文档。人像提示词的通用骨架从 300 案例中提炼的七要素通读portrait_ru.md全部案例后可以发现高质量人像提示词几乎都遵循同一套结构化写法。掌握这套骨架你就能把任意一个案例改写为自己的作品媒介与胶片风格声明开头一句话确立整体质感例如35mm film photography、Analog 35mm film photography、CCD camera aesthetic、mobile phone photo、DSLR photograph。这是决定成片像什么拍出来的最关键的一层。光源与曝光描述harsh direct on-camera flash机顶直闪、soft diffused natural window light柔和漫射窗光、mixed color temperature混合色温、golden hour黄金时刻等。身份与外貌细节在涉及真人/偶像的案例中会精确描写脸型、眼睛、鼻梁、下颌线、肤质visible pores、micro pores、妆容、发型。典型如 Case 1 中 ultra-realistic delicate refined Chinese features, seductive almond-shaped fox eyes with natural double eyelids, high nose bridge, small sharp V-shaped jawline, flawless porcelain skin with cool ivory undertone。服装与姿态衣物质地、褶皱、穿着方式oversized white button-up shirt、yukata deliberately slipped off one shoulder以及全身姿态与肢体语言body slightly arched、one leg bent。场景与环境便利店玻璃门、温泉旅馆木廊、篮球场黄昏、卧室床头……环境同时承担氛围与背景信息。后期与画质修饰词film grain胶片颗粒、no plastic skin、no digital over-sharpening、no airbrushing、no watermark、no text——大量案例在结尾用一长串否定词收束画质。输出参数--ar 9:16、--ar 2:3、--ar 4:5、--ar 3:4、--ar 1:1等宽高比控制以及部分案例给出的具体分辨率如1744x2336、1664x2080。下面的章节将按风格流派逐一展开并给出可直接复制的完整提示词。35mm 胶片与街头人像光影、颗粒与氛围的三层写法胶片风格是portrait_ru.md中出现频率最高的流派其提示词在媒介声明—光源—颗粒与色偏—场景氛围四层上做得极其细致。Case 1便利店霓虹胶片人像Convenience Store Neon Portrait这是典型的混合光源 真实胶片质感案例完整提示词如下35mm film photography with harsh convenience store fluorescent lighting mixed with colorful neon signs from outside, authentic film grain, high contrast, slight color cast, cinematic street editorial style, intimate medium shot, early 20s sexy Chinese female idol with ultra-realistic delicate refined Chinese features, seductive almond-shaped fox eyes with natural double eyelids, high nose bridge, small sharp V-shaped jawline, flawless porcelain skin with cool ivory undertone and visible specular highlights from fluorescent light, subtle skin texture and micro pores, natural dewy makeup with soft flush on cheeks, glossy natural pink lips slightly parted, subtle natural freckles across nose and cheeks, long dark brown hair in a messy high ponytail with many loose strands falling around face and neck, wearing an oversized white button-up shirt as the only top, unbuttoned at the top with deep cleavage and loosely tied at the waist, paired with a tiny black pleated mini skirt, barefoot in simple white slides, seductive casual leaning pose against the glass door of a 24-hour convenience store at late night, body slightly arched, one leg bent with foot resting against the door frame, the other leg straight, one hand holding a bottle of iced drink, the other hand lightly pulling the hem of her mini skirt, intensely seductive playful yet slightly vulnerable gaze straight at the viewer with soft doe eyes full of quiet temptation and teasing smile, bright cold fluorescent store light from inside mixed with pink and blue neon glow from outside signs, realistic reflections on glass door, blurred convenience store interior with shelves and snacks in background, authentic 35mm film color grading with harsh lighting and neon accents, extremely sharp yet soft skin rendering, natural hair strands, realistic fabric wrinkles and drape on the oversized shirt and mini skirt, no plastic skin, no digital over-sharpening, no airbrushing, no blemishes, no moles, no oily skin, no watermark, no text, authentic late-night convenience store atmosphere要点拆解媒介层用 35mm film photography authentic film grain slight color cast光源层明确 fluorescent lighting mixed with colorful neon signs 并规定色温方向 bright cold fluorescent ... mixed with pink and blue neon皮肤层是本案例最值得学习的地方——flawless porcelain skin 之后立刻补上 subtle skin texture and micro pores再以 no plastic skin, no digital over-sharpening, no airbrushing 收尾防止模型把完美皮肤渲染成塑料质感。这种先给形容词、再用否定词兜底的写法在整个文档中反复出现。Case 2极简电影人像Cinematic Minimal Portrait这是全文最短的高质量案例证明极简提示词同样有效Generate a cinematic minimal portrait of a solitary man standing in an intense orange to red gradient environment, strong silhouette lighting, deep shadow contrast, reflective glossy floor, symmetrical composition, minimal它只用五个视觉要素——颜色渐变环境、轮廓光、阴影对比、反光地面、对称构图——就定义了完整的画面。适合作为第一张测试图因为要素少任何参数调整换背景色、换人物姿态都能立刻看到效果差异。Case 4机顶直闪篮球场编辑人像35mm Flash Editorial Portrait直闪direct flash是胶片人像的另一大流派。本案例在结尾追加了宽高比参数35mm color film photography with harsh direct on-camera flash, specular highlights on skin and clothing, strong catchlights in eyes, high contrast flash illumination, authentic film grain and color shift, high fashion fresh innocent basketball court editorial style, intimate first-person low-angle POV shot from below, ... --ar 9:16完整提示词见 cases/portrait_ru.md Case 4注意 catchlights in eyes眼神光与 specular highlights镜面高光是直闪风格的专属关键词配合 harsh生硬一词还原真实机顶闪的物理特性。Case 6柔和空气感 35mm 人像Soft Airy 35mm Portrait与直闪相反这一派追求过曝与低对比Analog 35mm film photography, soft airy Japanese-style aesthetic, gentle diffused natural window light, slight overexposure, pastel tones, low contrast, soft highlights, minimal indoor setting near a window with white curtains, clean light-colored wall, natural composition, eye-level, slightly closer full-body framing (mid-thigh to head), young East Asian woman, natural minimal makeup, soft realistic skin texture, long slightly messy dark hair, oversized white button-up shirt, light casual shorts, barefoot, simple and relaxed styling, standing naturally with relaxed posture, arms loosely at sides or slightly behind, facing camera, gentle soft smile, subtle stillness, focus on light, air, and quiet everyday mood, soft film grain, dreamy and understated atmosphere --ar 9:16值得注意 focus on light, air, and quiet everyday mood 这类抽象氛围指令——它不描述具体物体而是给模型一个情绪方向与--ar 9:16参数结合是竖版日常人像的高频模板。身份一致性与多帧网格让同一张脸出现在九个画面GPT Image 2 的核心能力之一是跨生成保持角色一致。portrait_ru.md收录了大量同一人物多帧网格案例最典型的两种结构是 3x3 网格与 comp card模特卡。Case 11韩系偶像 3x3 网格Korean Idol 3x3 Grid Portrait极简版的一致性网格模板9:16 vertical, Korean idol portrait photoshoot, 3x3 grid (nine frames), same person in all images, consistent facial features and styling, soft black mist filter effect, lowered contrast, blooming highlights, subtle glow around light sources关键指令只有三条3x3 grid (nine frames)版式、same person in all images, consistent facial features and styling一致性约束、滤镜风格描述后期统一。Case 13带逐格分镜脚本的 3x3 拼贴Korean Idol 3x3 Collage Portrait进阶版把九宫格的每一格都写成明确分镜9:16 vertical — a 3x3 grid collage (nine images) forming a Korean idol portrait photoshoot series. Each frame features the same young Korean female idol, maintaining 100% consistency in facial features, proportions, hairstyle, and identity across all nine shots. ... Frame breakdown (3x3 grid): Top row: - Top left: standing naturally, looking slightly away, relaxed expression - Top center: facing camera, casual mid-motion ... Middle row: - Center left: looking slightly upward, soft thoughtful expression - Center: close-up portrait, direct eye contact, gentle idol smile ... Bottom row: - Bottom left: seated or leaning casually, relaxed posture - Bottom center: back partially turned, looking over shoulder toward camera - Bottom right: standing close to frame, slightly playful or soft expression ...完整分镜脚本见 cases/portrait_ru.md Case 13这个案例示范了一致性 分镜编排的组合先声明全局约束100% consistency、same outfit、same lighting再逐格给出动作、表情、景别差异。适用于需要一图多表情的偶像写真、社交媒体素材包等场景。Case 247黑白编辑风 comp cardMonochrome Editorial Comp Card同样基于 3x3 结构但用途转向模特卡模特经纪投放Ultra-realistic monochrome 3x3 editorial comp card using the reference image as the identity source. Preserve the exact same facial features, face shape, eye shape, nose, lips, skin texture, hairstyle, proportions, and natural beauty. No beautification, no face alteration, no AI-looking skin, no reshaping. Same woman in all 9 frames with consistent identity and styling. ... 3x3 grid layout with evenly spaced portrait frames. Each frame shows a different expression and hand pose: 1-looking sideways while lightly biting finger 2-side glance with hand touching temple ...除了逐格姿势外它还强调No beautification, no face alteration, no AI-looking skin——在商业场景中不美化与保持身份同等重要。类似的 2x3 六宫格版本见 Case 267 与 Case 271黑白编辑风 2x3 网格其中 Case 267 对发型hush cut、妆面、服装、六个分镜均有逐条描述是更精细的模板。参考图驱动的 UGC 工作流把照片变成可复用的模特portrait_ru.md中有一类方法论价值极高的案例——它们不只给提示词还给完整工作流。Case 21分析照片并生成 JSON 提示词Analyze and Recreate in JSON该案例演示了一条产品 UGC 素材流水线analyze this photo and give me a detailed JSON prompt that recreates it. break down the color grading and every exact color in the photo (use Opus, not Sonnet. Opus has stronger visual analysis and writes more detailed JSON) paste that JSON into ChatGPT upload your product image and prompt: using this JSON as reference, generate a person holding my product save that generated photo as your character reference attach it to every future generation for facial consistency you now have a consistent UGC model that works across any product the JSON controls the lighting and color grading. GPT image-2 handles the character. you control the product placement.工作流要点先用强视觉分析模型把参考照片的调色与逐色解析为 JSON再由 GPT Image 2 依据该 JSON 与产品图生成持产品的人物图最后把生成结果作为 character reference 反复复用。原文特意提示the #1 tell on AI photos is flat colors and a grainy look. this method removes both——即该流程同时解决 AI 照片色彩扁平与颗粒感失真两大破绽。参考图指令的两种强度在大量案例中可以观察到参考图uploaded image / reference image指令存在两种写法身份锁定型Use the uploaded portrait photo as the ONLY identity reference. Preserve face shape, facial features, hairstyle, skin tone, body proportion, and overall temperament. Do not copy the original photos background, lighting, clothing, pose, or image quality.参见 Case 183、Case 203 等——明确只取身份、不取场景是风格迁移场景下的标准写法。全身参照型Using REFERENCE_0 as the subject base, transform the casual desert snapshot into a full-body luxury fashion studio portrait. Replace the denim jacket, tank top, and shorts with ...Case 89——在保留身份的前提下允许替换服装与场景。杂志与印刷编辑风格从复古报纸头版到旅行指南页人像分类还包含大量印刷品设计案例把人物嵌进报纸、杂志、海报等版式系统。Case 19复古报纸头版Vintage Newspaper Front PageCreate the most realistic front page design of a vintage newspaper featuring the main character. The layout should be made in the style of a real printed newspaper with a cinematic black-and-white aesthetic. The main photo should be prominently placed in the center, framed, like the image in the title of the article. ... Add realistic newspaper elements: Columns of small text (in the style of lorem ipsum, but framed like real news) At the top is the fictitious name of the publication ... Date, issue number and location Decorative lines, dividers, and vintage typography ... Style: Black and white or slightly faded monochrome paper Fine paper texture, grain, and ink defects Small shadows and creases that mimic real printed paper ... Aspect ratio: 4:5 or 1:1该案例的启示在于版式类提示词要逐项列出版面元素标题、副标题、栏目文字、刊名、日期期号、装饰线、纸张纹理、折痕污渍并把人物照片作为版面主体嵌入其中。Case 20杂志旅行指南页Magazine Travel Guide Feature Article同类思路的杂志页版本核心是信息密度与文字准确性Create image of Magazine feature article [travel] guide page, cute, information dense photo book style magazine feature article page. Add all necessary sections, tips, recommendations, information. add photos for any sections and recommendations if you like. Place the attached person at the precise location of [city, country]. Seamlessly blend the attached person as if they are sightseeing. Approach this task with the understanding that this is a critical, information rich page that will significantly influence visitor numbers, text accuracy is important. Fully use the entire [9:16] page. NEGATIVE PROMPT: coordinate texts swiat_ai ProfitAII注意其中 text accuracy is important 的显式强调——GPT Image 2 具备较高保真的文字渲染能力README 中亦将其列为关键能力 High fidelity text rendering因此在杂志、海报类提示词中主动声明文字必须准确能显著提升成片可用性。跨风格迁移与插画化把照片变成另一种媒介风格迁移style transfer是portrait_ru.md另一大类覆盖蜡笔涂鸦、木刻版画、线稿、水彩、折纸、毛毡、Pixar 3D 等多个方向。儿童蜡笔风Case 31 / Case 255in the style of super bad child drawing, toddler art, scribbles, messy crayon lines on white background, completely lack of technique, terrible composition, chaotic colors, barely recognizable shapes ...配套负向提示good drawing, nice lines, clear shapes, neat, pretty, smooth, realistic, talented art ...。通过反向修辞描述糟糕的技法实现幽默风格。木刻版画Case 205Black and white engraved portrait illustration ... classic woodcut / linocut engraving style, high contrast black ink on textured off-white paper background. Fine cross-hatching and line shading ...负向提示明确排除color, watercolor, soft shading, realistic photography, 3D render, anime style。线稿与水彩混合Case 213A precise line art illustration portrait in the distinctive Miki style, with clean, defined dark contours and minimal, subtle watercolor wash coloring.折纸艺术Case 176Ultra-detailed origami paper art portrait of given picture, entirely crafted from meticulously folded paper layers and intricate geometric origami shapes. Realistic paper texture with visible creases and handcrafted folds ...羊毛毡微缩Case 304Transform the subject into a handcrafted needle-felted wool miniature. Material: organic roving wool with visible needle-punch textures, soft fuzzy surface, and handcrafted seams.Pixar 3D 角色设计稿Case 206 / Case 236Create a Pixar 3D style character design sheet. Clean white background. Two characters side by side with a clean dividing line. Bold brushstroke-style title at the top ...这类案例的共同方法论是媒介材料先行、工艺细节随后先声明目标媒介蜡笔、版画、羊毛毡再用该媒介的物理特性词针毡纹理、排线、折痕、纸纤维约束渲染结果最后用负向提示排除照片级真实感。负向提示词与画质控制高频否定词的语义归纳portrait_ru.md中大量案例自带 Negative Prompt负向提示。对全文否定词做词频归纳可归为四类类别高频词示例用途皮肤与人物plastic skin、waxy skin、over-smoothed skin、beauty filter、doll face防止塑料感 AI 脸结构与解剖extra fingers、missing fingers、deformed hands、bad anatomy、distorted face、asymmetrical eyes防止手指、五官、肢体畸形风格污染cartoon、anime、3D render、CGI、illustration防止偏离写实目标画面杂质watermark、logo、text、noisy image、low resolution、blurry face防止水印、文字、噪点几个可直接复用的完整负向提示模板Case 234通用模板low quality, blurry, distorted face, extra fingers, bad hands, duplicate facial features, unrealistic reflection, cartoon, anime, watermark, logo, text, noisy image, oversaturated colors, poorly drawn eyes, bad anatomy, broken proportions, low resolution.Case 104写实美妆向over-smoothed skin, plastic texture, unrealistic proportions, studio lighting, overly sharp HDR, stiff pose, artificial symmetry, over-retouched face.Case 300自拍写实向low quality, blurry, distorted face, bad anatomy, extra limbs, stiff pose, unnatural selfie angle, overexposed skin, harsh flash, plastic skin, overly bright colors, cheap fabric, messy background, cartoon style, exaggerated beauty filter, unrealistic eyes, artificial hair, bad hands, awkward arm, noisy image.值得注意负向提示并非独立参数而是作为正文的一部分写进同一段提示词如 Case 25 在提示词末尾以 Negative Prompt: 分段因此复制案例时需连同分隔符一并保留。宽高比与分辨率参数--ar 与显式尺寸文档各案例中出现的输出尺寸控制方式有两种--ar语法仿 Midjourney 风格见 Case 4、Case 6、Case 112、Case 114、Case 211、Case 310 等--ar 9:16竖版手机全屏偶像写真、自拍、短视频封面高频使用--ar 2:3标准竖版人像--ar 4:5社交平台通用竖版Instagram 常用--ar 3:4兼容老式竖幅照片--ar 1:1方形网格、报纸头版显式分辨率部分案例直接给出像素尺寸如 Case 25 的1664x2080-ar 4:5、Case 198 与 Case 230 的1744x2336。这类写法将尺寸直接写进画面描述适合对构图空间有精确要求的版式类任务。在英文 README.md 的模型能力说明中Native support for aspect ratios, transparency, and batch generation 被列为 GPT Image 2 API 的原生能力因此宽高比参数可以直接透传到 API 请求。把这些提示词跑起来从案例库到 API 调用仓库提供了把上述任意提示词转化为真实图片的最小路径。以英文 README.md 的 EvoLink 快速开始为例安装技能后只需一条 curl 命令export EVOLINK_API_KEYyour_key_here curl --request POST \ --url https://api.evolink.ai/v1/images/generations \ --header Authorization: Bearer ${EVOLINK_API_KEY} \ --header Content-Type: application/json \ --data { model: gpt-image-2, prompt: A premium ecommerce hero shot of a ceramic coffee dripper on a clean kitchen counter, soft morning light, realistic product photography }把prompt字段替换为本文任一完整提示词即可生成对应人像。仓库还提供npx evolink-gpt-image -y一键安装的 Agent Skill详见 README.md 的 Use GPT Image 2 API 章节支持将仓库内提示词直接编排为可调用技能。多语言 README如 README_ru.md与案例文档如 cases/portrait_ru.md可通过 script/sync_multilingual_readmes.py 保持同步更新。案例如何进入仓库从社区发现到收录人像文档不是一次性整理的结果而是持续运营的产物。仓库data/目录下的 curation report如 data/curation_report_2026-06-08.json记录了每日案例收录流水线原始抓取 → 去重 → 预筛选 → 人工评审 → 推荐收录并给出每个候选案例的推文链接、作者、点赞数、浏览量、提取出的提示词、建议标题与分类。英文 README.md 的 News 部分也持续记录着每周新增的人像案例数量如 June 18, 2026: Added ... 2 portrait ...。这意味着本文拆解的所有模板都在持续扩充当你阅读本指南时portrait_ru.md与其英文源文件 cases/portrait.md 已经可能收录了更新的案例。任何提示词改动都遵循同一套结构化书写 逐格分镜 负向收尾的方法论本文提炼的骨架不会过时。小结从案例库到你的第一张人像cases/portrait_ru.md的价值不在于某个单独的提示词而在于它提供了完整的人像提示词谱系35mm 胶片与混合光源的物理质感、身份一致性的多帧网格编排、参考图驱动的 UGC 流水线、印刷版式的元素罗列、跨媒介风格迁移的工艺描述以及贯穿始终的负向提示词纪律。上手路径建议先跑 Case 2 的极简电影人像验证 API 与宽高比参数挑一个你喜欢的流派霓虹胶片或 CCD 闪光逐词拆解其光源与皮肤层写法需要多图素材时直接套用 3x3 网格模板并自定义分镜有真实产品需求时按 Case 21 的 JSON 工作流建立自己的 UGC 模特参考。所有案例的完整提示词、输出图片与原始出处均可直接查阅 cases/portrait_ru.md 及对应图片目录如 images/portrait_case1、images/portrait_case2按图索骥即可。【免费下载链接】awesome-gpt-image-2-API-and-PromptsGPT-Image-2 API and Prompts项目地址: https://gitcode.com/gh_mirrors/aw/awesome-gpt-image-2-API-and-Prompts创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考