PROMPT RADAR / PROMPT人像时尚

YOU HAVE BEEN TRAINED TO TRUST A PHOTO YOUR OWN EYES CAN NO LONGER VERIF

YOU HAVE BEEN TRAINED TO TRUST A PHOTO YOUR OWN EYES CAN NO LONGER VERIFY That queso smash on the sign outside? It was never plated, never cooked, never real. It was generated by a prompt, approved by a manager who thoug

@grokkedd 481 热度 3.4 万 浏览
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。 先锁定人物身份与服装轮廓,再用光线、焦段和动作消除“普通写真感”。

完整视觉 Prompt

YOU HAVE BEEN TRAINED TO TRUST A PHOTO YOUR OWN EYES CAN NO LONGER VERIFY That queso smash on the sign outside? It was never plated, never cooked, never real. It was generated by a prompt, approved by a manager who thought nobody would notice. > menu boards, delivery apps, and social platforms are increasingly running AI-generated food photography instead of real shots - Forkable, a US catering platform, quietly swapped real restaurant photos for AI images across its entire site without telling the restaurants themselves > McDonald's Netherlands pulled a fully AI-generated Christmas ad within days of backlash, proof that even a billion-dollar brand can't out-market the moment customers spot the uncanny gloss of a synthetic image > under the FTC's "reasonable consumer standard," AI food imagery can legally count as false advertising the moment it materially changes what someone expected to receive - a disclaimer buried in fine print doesn't erase that > the deception now runs both directions: some customers use AI-edited photos of "raw" or "moldy" food to fraud their way into refunds, forcing DoorDash and other platforms to build detection systems just to catch fabricated evidence > researchers call this a live regulatory blind spot - almost no jurisdiction has updated food fraud law to cover AI-generated imagery, meaning there is currently very little legal deterrent for either the restaurant or the customer How to actually spot it before you order: > look for unnaturally perfect symmetry - real melted cheese pulls unevenly, AI cheese pulls in a suspiciously smooth arc > check the shadows - AI food photos often have lighting that doesn't match the direction of any visible light source > zoom in on garnish and texture - AI still struggles with fine details like sesame seed placement, herb sprigs, or steam > search the exact dish name plus the restaurant on Reddit or Instagram - real customer photos will contradict the marketing fast You were taught to trust your eyes at the counter. Now the eyes need training too.
查看原作者内容

不要一次改完所有变量

用 Grok Imagine 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:YOU HAVE BEEN TRAINED TO TRUST A PHOTO YOUR OWN EYES CAN NO LONGER VERIFY That queso smash on the sign outside? It was never plated, never cooked, never real. It was generated by a prompt, approved by a manager who thoug
  2. 视觉方向:实时 AI 视觉案例;查看原帖媒体与作者说明
  3. 镜头与构图:查看原帖画面;查看原帖媒体与工作流
  4. 交付约束:auto,保留原作者方法与归属;复刻时避免水印、低清、畸变、主体漂移和无关文字。
负向约束

保留原作者方法与归属;复刻时避免水印、低清、畸变、主体漂移和无关文字。

模型建议

Grok Imagine / GPT Image 2 / 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:grokked · 收录于 2026年7月14日

访问原始来源

沿着这条线继续看