原始文档
GPT Image 2 · en
Scene:
Compose a square brunch tabletop image with layered tableware, generous daylight, and a sense of weekend abundance that still feels organized enough for real hospitality and publishing use. The scene should balance multiple objects without losing the hero focal point, making table choreography and dish relationships a core part of the image. Use a square overhead-to-three-quarter view with warm daylight, a woven basket, small ramekins, and linen folds that feel casual but polished.
Subject:
The food must look fresh, warm, and genuinely shareable: eggs, breads, fruit, spreads, juices, and plated brunch mains should feel cooked and plated for service, not arranged as fake prop food. Feature croissants, rolls, butter, preserves, and one or two plates nearby so the basket reads as part of a brunch service rather than an isolated still life.
Important details:
Table logic matters. Plates, glasses, linens, and serving utensils need believable spacing, overlap, and hierarchy so the image can support menus, social assets, and campaign layouts. Make the abundance feel intentional rather than chaotic, with appetizing textures and natural daylight carrying the mood. This image should feel fully self-contained while still belonging to the same broader topic and commercial world. Keep textures rich but disciplined: flaky pastry, soft butter, glass jar reflections, and cloth folds should feel inviting without becoming cluttered.
Use case:
Brunch Tabletop single study 5/9, built for Prompt Atlas coverage of a food and beverage commercial visual system.
Constraints:
No readable text, no watermark, no logos, preserve exact 1:1 aspect ratio, no buffet chaos, no greasy diner styling unless specifically implied, no visible phone screens, no plastic takeaway packaging, no fake stock-family posing, no unreadable signage, no supermarket pastry pile, no plastic packets, no buffet sneeze guards.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。
