GPT Image 2 推理海报
一张光泽感 3D 社交媒体海报，展示了一个立方体大脑图标，并配有醒目的文案，解释了 GPT Image 2 的“规划-搜索-生成-验证”工作流程。
Goal: Create a polished vertical promotional poster for {argument name="product name" default="GPT Image 2"} explaining that the image model thinks before rendering, using a playful 3D toy-like visual style.
Canvas: 4:5 vertical social media poster, deep cobalt-blue background with a subtle radial glow and soft vignette, high-resolution glossy 3D render aesthetic.
Main visual: Center the top half around a surreal 3D object: a rounded glossy royal-blue cube fused with a shiny pink brain hemisphere protruding from the cube's right side. The cube is angled slightly so the front-left face and top are visible, with soft studio highlights, rounded edges, and a grounded shadow. The brain should look like inflated candy or silicone, with rounded folds, bright pink highlights, and a soft magenta glow where it meets the cube.
Decorative elements: Scatter exactly 34 small floating candy/confetti objects around the poster, avoiding the main text area. Include these visible types: 1 yellow star, 2 curved jellybean shapes, 11 short capsule sprinkles, 11 small spheres/dots, 4 larger glossy balls, 3 tiny round dots, and 2 partial edge sprinkles. Use pink, yellow, mint green, purple, orange, white, and red, with soft shadows and glossy highlights.
Text layout: Use bold white sans-serif typography, centered. Large headline in the middle: “{argument name="headline text" default="It reasons, then renders."}” on two lines, very heavy weight, clean geometric sans serif, slight shadow for contrast. Beneath it, smaller bold subtitle: “{argument name="subtitle text" default="GPT Image 2 plans your image before it makes a pixel."}” on two centered lines.
Instruction card: At the lower third, place a rounded off-white command-card panel with a subtle drop shadow, like a copyable prompt box. On the right side, attach a coral-red rounded rectangular button partially behind the card, labeled “COPY” in dark bold text. Inside the card, show exactly 4 monospaced lines in dark charcoal: “PLAN: composition and constraints.”, “SEARCH: facts you need, or SKIP.”, “GENERATE: the brief.”, “VERIFY: check output before showing.”
Footer: Bottom-left, add a tiny white line-art dog/mascot icon followed by the text “{argument name="footer label" default="GPT IMAGE 2 · PLAIN AND SIMPLE"}” in small uppercase white sans serif. Bottom-right, add the progress text “{argument name="progress text" default="1 / 7"}”.
Visual style: Bright modern 3D advertising, glossy plastic materials, candy colors, rounded friendly forms, strong contrast, crisp typography, soft realistic shadows, no photorealistic people, no watermark, no extra text.目标：为 {argument name="product name" default="GPT Image 2"} 创建一张精致的竖版宣传海报，通过俏皮的 3D 玩具风格视觉效果，解释该图像模型在渲染前会进行思考。
画布：4:5 竖版社交媒体海报，深钴蓝色背景，带有微妙的径向光晕和柔和的晕影，呈现高分辨率光泽感 3D 渲染美学。
主视觉：上半部分中心是一个超现实的 3D 物体：一个圆润且富有光泽的宝蓝色立方体，右侧嵌入一个闪亮的粉色大脑半球。立方体略微倾斜，露出左前侧和顶部，带有柔和的摄影棚高光、圆角和落影。大脑应看起来像充气的糖果或硅胶，具有圆润的褶皱、明亮的粉色高光，以及在与立方体连接处散发的柔和洋红色光芒。
装饰元素：在海报周围散布 34 个小的悬浮糖果/五彩纸屑物体，避开主要文字区域。包括以下可见类型：1 个黄色星星、2 个弯曲的软糖形状、11 个短胶囊状糖粒、11 个小球体/圆点、4 个较大的光泽球体、3 个微小圆点和 2 个边缘碎片。使用粉色、黄色、薄荷绿、紫色、橙色、白色和红色，并带有柔和的阴影和光泽高光。
文字排版：使用居中的粗体白色无衬线字体。中间的大标题：“{argument name="headline text" default="It reasons, then renders."}”分为两行，采用极粗的字重、简洁的几何无衬线字体，并带有轻微的阴影以增强对比度。下方是较小的粗体副标题：“{argument name="subtitle text" default="GPT Image 2 plans your image before it makes a pixel."}”，分为两行居中排列。
指令卡片：在下方三分之一处放置一个圆角灰白色命令卡片面板，带有微妙的投影，类似于可复制的提示词框。在卡片右侧附带一个珊瑚红色的圆角矩形按钮，部分位于卡片后方，上面用深色粗体文字标注“COPY”。卡片内显示 4 行深炭灰色的等宽字体：“PLAN: composition and constraints.”、“SEARCH: facts you need, or SKIP.”、“GENERATE: the brief.”、“VERIFY: check output before showing.”
页脚：左下角添加一个微小的白色线条艺术风格狗/吉祥物图标，后接小号大写白色无衬线文字“{argument name="footer label" default="GPT IMAGE 2 · PLAIN AND SIMPLE"}”。右下角添加进度文字“{argument name="progress text" default="1 / 7"}”。
视觉风格：明亮的现代 3D 广告风格，光泽塑料材质，糖果色调，圆润友好的形态，强烈的对比度，清晰的排版，柔和逼真的阴影，不包含写实人物，无水印，无额外文字。