PROMPT RADAR / IMAGE平面设计
个人资料 / 头像 - Cinematic Red Blue Split-Light Portrait
Generates a dramatic ultra-realistic studio portrait with red and blue split lighting for premium editorial or social media visuals.
@abs_uiux 0 热度 0 浏览
ORIGINAL VISUAL / 来源见文末编辑先看
它为什么值得进入灵感库
它最有价值的部分,是把一个模糊想法推进到了可以被模型执行的程度。 先规定信息层级和文字安全区,再谈视觉风格;无法落版的漂亮画面不是完整设计。
01 / 原始方法
完整视觉 Prompt
Create an ultra-realistic cinematic close-up studio portrait of a {argument name="subject" default="confident young adult Black man"} facing the camera directly, centered in the frame against a completely dark black background. He has a short neat fade haircut, smooth skin with natural pores and subtle highlights, dark focused eyes, relaxed closed lips, and a calm powerful expression. Pose him with one hand thoughtfully under his chin, thumb along the jawline, showing a polished silver metal wristwatch with a black dial on the wrist. Dress him in a plain {argument name="shirt color" default="black"} crew-neck T-shirt. Use dramatic split lighting: saturated electric {argument name="left light color" default="blue"} light illuminating the viewer-left side of his face, hand, and shoulder, and saturated {argument name="right light color" default="red"} light illuminating the viewer-right side, with a subtle blend across the center of the face. Make the image feel premium, masculine, mysterious, editorial, and powerful, like a high-end magazine cover portrait. Use tight head-and-shoulders framing, shallow depth of field, crisp facial detail, realistic skin texture, glossy catchlights in the eyes, strong contrast, deep shadows, no visible text, no logos, no watermark, no props besides the watch, and no background detail.
02 / 复刻路径
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。
- 主体:Generates a dramatic ultra-realistic studio portrait with red and blue split lighting for premium editorial or social media visuals.
- 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
- 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
- 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