GPT Image 2 一致性角色设定表的提示词：

The prompt for the consistent character sheet with GPT Image 2： Create a new original 2K 16:9 character consistency sheet for the same fictional lifestyle creator from the reference image. Preserve her core identity, hair color, outfit concept, KOL fashion mood, and accessories, but redraw the whole sheet as a clearly synthetic high-end 3D digital human, not a live-action photo and not a real person. Visual direction: stylized-realistic 3D character art, cinematic game/animation quality, realistic proportions and cloth folds, but visibly CG: clean shader skin, sculpted facial planes, controlled pores, groomed hair strands, polished virtual studio lighting. Real enough for a vlog, but obviously designed and rendered. Subject: fictional young Caucasian / Mediterranean-Caucasian woman, warm sun-kissed fair skin, green-hazel eyes, honey chestnut-blonde shoulder-length waves, small gold earrings, calm friendly KOL expression. Outfit: sage-green relaxed utility jacket, cream ribbed knit top, sand-khaki wide-leg trousers, black high-top sneakers, tan structured handbag, delicate jewelry, plain lidded takeaway coffee cup. Layout: clean wide studio model sheet. Large full-body FRONT, SIDE, BACK views on the left; smaller 3/4 angle and expression panels on the right. English labels only: FRONT, SIDE, BACK, 3/4 VIEW, PROFILE, NEUTRAL, SMILE, CURIOUS, DETAILS. Keep face close-ups small and secondary; emphasize full body, outfit, silhouette, bag, shoes, and coffee cup. Quality: same woman and outfit in every panel, consistent body proportions, crisp garment texture, beige-gray background, thin dividers, professional digital character sheet. Avoid: photoreal live-action portrait, real model photo, biometric face scan, passport/ID layout, huge isolated face panels, skin pore macro, facial-part closeups, stock-photo lighting, phone photo look, logos, watermark, QR code, Chinese/Japanese/Korean text, random symbols, anime, cartoon, plastic doll face, distorted hands, duplicate limbs, bad cup, misspelled title.

GPT Image 2 一致性角色设定表的提示词： 基于参考图，为同一位虚构的生活方式博主创建一张全新的 2K 16:9 角色一致性设定表。保留她的核心身份、发色、穿搭理念、KOL 时装气质与配饰，但将整张表重新绘制为明显合成的、高品质的 3D 数字人——不是真人照片，也不是真实人物。 视觉方向：风格化写实 3D 角色美术，电影级游戏/动画品质，真实的比例与衣褶，但明显是 CG：干净着色器皮肤、雕刻般的面部平面、受控的毛孔、梳理整齐的发丝、打磨过的工作室虚拟灯光。足够真实，可以当作 vlog 画面，但显然是设计与渲染出来的。 主体：虚构的年轻高加索 / 地中海-高加索女性，温暖阳光感的白皙肤色，绿色-榛色眼眸，蜂蜜栗金色齐肩波浪卷发，小巧金色耳环，神情平静友好的 KOL 表情。服装：鼠尾草绿色宽松工装夹克，奶油色罗纹针织上衣，沙卡其色阔腿长裤，黑色高帮运动鞋，棕褐色结构感手提包，精致首饰，一只素色带盖外带咖啡杯。 版式：干净宽阔的工作室模特设定表。左侧为大幅全身 FRONT（正面）、SIDE（侧面）、BACK（背面）视图；右侧为较小的 3/4 角度与表情小图。仅使用英文标注：FRONT、SIDE、BACK、3/4 VIEW、PROFILE、NEUTRAL、SMILE、CURIOUS、DETAILS。面部特写保持小且次要；重点突出全身、服装、廓形、包、鞋和咖啡杯。 质量：每个面板中保持同一位女性与同一套服装，身体比例一致，服装纹理清晰，米灰背景，细分割线，专业级数字角色设定表。 避免：写实真人肖像、真实模特照片、生物识别面部扫描、护照/证件版式、超大独立面部面板、皮肤毛孔微距、面部五官特写、图库照片打光、手机拍照观感、标志、水印、二维码、中文/日文/韩文、随机符号、动画、卡通、塑料娃娃脸、手部变形、重复肢体、劣质杯子、标题拼写错误。