GPT Image 2 on ChatGPT

GPT Image 2 on ChatGPT UPLOAD A REFERENCE IMAGE BEFORE GENERATING. MAIN PROMPT Transform every visible person in the uploaded image into a realistic LEGO-style minifigure while preserving the original environment, background, composition, lighting, and relationships. Create exactly one separate minifigure for every visible person. Preserve each person’s recognizable identity, expression, hairstyle, hair color, skin tone, proportions, pose, position, clothing, accessories, and held objects. FORMAT LOCK Vertical 3:4 composition. Show every minifigure fully from head to toe without cropping. Preserve the original group arrangement and perspective. Extend the environment naturally only when needed for the 3:4 frame. STYLE TRANSFER Convert only the people into authentic LEGO-style minifigures. Use realistic molded plastic, cylindrical heads, block-shaped torsos, standard arms and hands, short legs, printed facial features, detailed printed clothing, and molded plastic hairpieces matching each person’s hairstyle. Keep the figures seamlessly integrated into the original photograph. DETAIL & COLOR Accurately reproduce each person’s original clothing colors, patterns, layers, sleeves, collars, pants, skirts, shoes, jewelry, glasses, hats, bags and other visible accessories as LEGO prints or molded elements. Preserve original skin tones and facial expressions. Realistic plastic reflections, subtle seams, fine printed details and natural contact shadows. COMPOSITION LOCK Keep the original location, architecture, scenery, furniture, objects, colors, textures, perspective, depth, weather, atmosphere and time of day unchanged. Do not redesign or simplify the background. Preserve all existing objects in their original positions. Objects held or worn by people may be converted into LEGO-style accessories while maintaining their original appearance and color. STYLE Photorealistic LEGO-style minifigure transformation, premium 3D toy photography, realistic molded plastic, subtle glossy finish, physically accurate reflections, natural lighting, cinematic detail, seamless integration with the original scene. NEGATIVE PROMPT No changed background, studio backdrop, LEGO environment, added bricks, removed objects, extra people, missing people, duplicates, merged figures, generic characters, altered outfits, invented accessories, identity loss, wrong hairstyles, cropped bodies, missing limbs, oversized heads, realistic human bodies, real hair, real fabric, anime, cartoon exaggeration, CGI-looking environment, floating figures, excessive reflections, harsh shadows, blur, low resolution, text, logos, watermark.

GPT Image 2 on ChatGPT 在生成之前上传一张参考图像。 主提示词 将上传图片中每一个可见的人物转换为逼真的乐高风格小人仔，同时保留原始环境、背景、构图、光照和人物关系。为每一个可见人物精确创建一个独立的小人仔。保留每个人物可辨认的身份、表情、发型、发色、肤色、比例、姿势、位置、服装、配饰以及手持物品。 格式锁定 竖向 3:4 构图。完整显示每个小人仔从头到脚的全身，不进行裁切。保留原始的群体排列和透视。仅在 3:4 画框需要时，自然延伸环境。 风格转换 仅将人物转换为正宗的乐高风格小人仔。使用逼真的注塑塑料、圆柱形头部、块状躯干、标准手臂与手部、短腿、印制的面部特征、精细印制的服装，以及与每个人物发型匹配的注塑塑料发饰。让人物与原始照片无缝融合。 细节与色彩 通过乐高印花或注塑零件，准确再现每个人物原有的服装颜色、图案、层次、袖子、领口、裤子、裙子、鞋子、珠宝、眼镜、帽子、包以及其他可见配饰。保留原有的肤色和面部表情。呈现逼真的塑料反光、细微的接缝、精致的印花细节以及自然的接触阴影。 构图锁定 保持原始的地点、建筑、景观、家具、物体、色彩、纹理、透视、深度、天气、氛围和时间段不变。不要重新设计或简化背景。保留所有现有物体的原始位置。人物手持或佩戴的物体可以转换为乐高风格的配件，同时保持其原有的外观和颜色。 风格 照片写实的乐高风格小人仔转换，高端 3D 玩具摄影，逼真的注塑塑料，微妙的亮面质感，物理上准确的反光，自然光线，电影级细节，与原始场景无缝融合。 负面提示词 不要改变背景、不要工作室背景、不要乐高环境、不要添加积木、不要移除物体、不要多出人物、不要缺失人物、不要重复、不要合并人物、不要通用角色、不要改动服装、不要虚构配饰、不要丢失身份、不要错误发型、不要裁切身体、不要缺少四肢、不要头部过大、不要逼真的人体、不要真实头发、不要真实布料、不要动漫、不要卡通夸张、不要 CGI 风格的环境、不要漂浮人物、不要过度反光、不要刺眼阴影、不要模糊、不要低分辨率、不要文字、不要 logo、不要水印。