
原始文档
GPT Image 2 · en
Scene:
Compose a square food-commerce image optimized for instant readability in app tiles, promo cards, or delivery grids. The image should emphasize appetizing clarity, strong hierarchy, and modular layout zones while staying photorealistic and commercially grounded. Packaging, dish crop, and badge-like design shapes can appear, but all text must remain unreadable. Preserve the same card geometry and high-key commerce lighting, but add one extra side or beverage cue to imply a bigger offer.
Subject:
The food should look craveable at thumbnail scale, with clear portion logic, supporting side items when relevant, and delivery-aware presentation that prioritizes speed, legibility, and conversion. Feature a main dish with one clearly visible add-on and organized packaging so the tile communicates a more substantial combo without crowding the frame.
Important details:
This theme is about commerce reasoning. Preserve sharp hero food silhouettes, obvious combo structure, and clean tile-safe framing so the assets can survive small screens. Abstract text areas, CTA-like shapes, and offer badges are allowed only as non-legible design modules. Keep everything polished, modern, and immediately useful for ordering surfaces. Preserve the exact camera height, lens feel, prop family, tabletop scale, palette discipline, and brand logic established by 01-app-tile-anchor. The anchor hierarchy must remain intact: hero food first, support item second, abstract badges third. Keep everything crisp enough for mobile browsing.
Use case:
Delivery Tiles series study 2/9, 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 readable text, no logos, no app UI chrome, no fake prices, no delivery-bag chaos, no greasy blur, no clutter that hurts thumbnail legibility, no readable promo text, no huge feast pile, no delivery driver props.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。