PROMPT RADAR / PROMPT插画三维

社交媒体帖子 - Bird's-Eye Street Fashion Portrait

A detailed streetwear fashion prompt for gpt-image-2 that utilizes reference images for scene composition and facial identity while randomizing clothing.

@AiwithLariab 0 热度 0 浏览
社交媒体帖子 - Bird's-Eye Street Fashion PortraitORIGINAL VISUAL / 来源见文末
编辑先看

它为什么值得进入灵感库

先别急着照抄整段 Prompt:真正有用的是主体、空间和镜头之间的约束关系。 轮廓语言、材质规则和世界观要互相支持,避免把互不相干的风格词堆在一起。

完整视觉 Prompt

ULTRA-REALISTIC CINEMATIC STREET FASHION PORTRAIT — 8K, RAW DSLR QUALITY, MASTERPIECE, HYPER-DETAILED, PHOTOREALISTIC. Use the FIRST UPLOADED IMAGE as the MAIN SCENE, CAMERA ANGLE, POSE, COMPOSITION, BACKGROUND, AND BODY POSITION REFERENCE. Use the SECOND UPLOADED IMAGE as the EXCLUSIVE FACIAL IDENTITY REFERENCE. A {argument name="subject" default="young Asian woman"} standing casually on a wide asphalt road with bold alternating black-and-white painted stripes. Photographed from a high overhead/bird’s-eye perspective, looking directly up toward the camera with a calm, confident, slightly cool expression. Her body is relaxed, both hands naturally inside her pants pockets, with one leg slightly bent and crossed casually in front of the other. Preserve the exact overall pose, perspective, framing, and visual composition of the first reference image. FACE & IDENTITY — CRITICAL: Use the SECOND uploaded photo to reproduce the person's facial identity as accurately as possible. Preserve her exact facial structure, face shape, forehead, eyebrows, eyes, eyelids, nose shape, lips, cheeks, jawline, chin, ears, natural skin texture, complexion, and natural facial proportions. Do not replace, redesign, beautify, masculinize, feminize, or stylize her face. The final face must clearly remain the same person from the facial reference. HAIR: The hairstyle must follow the person in the SECOND uploaded facial reference. Preserve her natural hair characteristics, hairline, texture, thickness, length, volume, and overall hairstyle. Do not copy the hairstyle from the first scene reference. OUTFIT: Generate a {argument name="style" default="RANDOM feminine modern streetwear outfit"} that is completely different from the clothing in both reference images. The outfit can be an oversized T-shirt, relaxed-fit shirt, hoodie, bomber jacket, varsity jacket, denim jacket, casual blazer, or other stylish contemporary streetwear piece, paired with relaxed or loose-fit pants and fashionable sneakers. Randomize the clothing colors, patterns, and design while keeping the entire outfit cohesive, realistic, feminine, trendy, stylish, and effortless. ACCESSORIES: Random subtle feminine accessories are allowed, such as a watch, bracelet, simple rings, small earrings, or a minimal necklace. Keep accessories natural, understated, and realistic. LIGHTING & VISUAL STYLE: Natural outdoor daylight, soft realistic shadows, authentic asphalt texture, realistic fabric texture, natural skin tones, cinematic color grading, subtle film grain, realistic depth and perspective, premium street-fashion photography aesthetic, RAW DSLR quality. CAMERA & COMPOSITION: High-angle bird’s-eye/top-down photography, camera positioned directly above the subject, full-body shot, vertical 9:16 composition, subject centered in frame, realistic head-to-body proportions, natural anatomy. Preserve the strong geometric black-and-white road stripes from the first reference image.
查看原作者内容

不要一次改完所有变量

用 GPT Image 2 复刻时,先固定主体和构图,再逐次替换光线、色调或材质中的一个变量,这样更容易判断哪项描述真正起作用。

  1. 主体:A detailed streetwear fashion prompt for gpt-image-2 that utilizes reference images for scene composition and facial identity while randomizing clothing.
  2. 视觉方向:社区实测摄影提示词;按原提示词的光线描述执行
  3. 镜头与构图:按原提示词的镜头与景别执行;按原提示词的构图与画幅执行
  4. 交付约束:auto,遵循原提示词中的避免项;不要水印、乱码、低清、畸形手、错误透视或主体漂移。
负向约束

遵循原提示词中的避免项;不要水印、乱码、低清、畸形手、错误透视或主体漂移。

模型建议

GPT Image 2 / Grok Imagine / Midjourney / Flux。先用原始 Prompt 建立基准结果,再根据模型对自然语言、镜头运动或文字排版的能力做局部改写。

来源与编辑原则

保留作者,也保留判断。

本页收录公开案例并提供中文索引、结构化拆解和复刻建议,不冒充原作者作品,也不替代原帖上下文。

作者:Laraib Fatima‎ · 收录于 2026年8月10日

访问原始来源

沿着这条线继续看