
原始文档
GPT Image 2 · en
Scene:
A square premium-streaming key art composition staged inside a glass-walled penthouse above a winter city skyline at blue hour. The frame is cool, polished, and austere, with silver reflections, pale marble, smoked glass, and a disciplined balance of shadow and controlled highlights. The camera is frontal and slightly elevated so the whole group reads like an official campaign image rather than a candid scene.
Subject:
Create a fictional prestige-drama ensemble of six people arranged in a strict power pyramid. At the center sits an older matriarch in an ivory suit on a low modern chair, perfectly still and clearly dominant. Behind her stands the chosen heir in a dark charcoal suit, tense but composed. To one side, a sharp younger daughter leans against the window with a guarded stare; opposite her, an outsider son-in-law stands half a step back, uncertain but ambitious. Slightly behind them, a discreet fixer and a silent bodyguard complete the composition. Their expressions must suggest inheritance battles, private grievances, and long family memory without melodrama.
Important details:
This is the reference image for the series, so face identities, hair silhouettes, wardrobe palette, and precise spacing between figures must be memorable and reusable. The visual language must feel like an awards-season streamer campaign: clean negative space, expensive tailoring, subtle environmental storytelling, and unmistakable role hierarchy. Preserve believable anatomy, consistent eye lines, and enough empty space around the ensemble for future crop logic.
Use case:
Prestige TV Key Art Remix series study 1/9, establishing the core cast hierarchy for a fictional streaming dynasty drama.
Constraints:
No readable text, no watermark, no logos, preserve exact 1:1 aspect ratio, no comedy energy, no crowded props, no extra background people, no exaggerated action poses, no cheap soap-opera styling.使用方法
- 复制完整原始提示词,并保留其中的参数。
- 按需求替换主体、文字和构图要求。
- 在兼容的外部工具中使用,核对当前使用条件并检查结果。
开始前需了解
- 先确认是否需要参考图。标注兼容模型,不代表该示例由这个模型生成。
- 核对名称、数字与画面文字;不同模型版本的结果可能不同。





