
原始文档
GPT Image 2 · en
Scene:
A square artifact and material study arranged like a high-end conservation photograph. Use soft directional light, controlled shadow falloff, and a neutral work surface so every fragment reads clearly. The image should feel intimate and scientific, but still rich enough to communicate the broader civilization through craft, wear, and material vocabulary alone.
Subject:
Present jade beads, obsidian blades, shell inlay, pottery shards, and carved limestone fragments arranged on a padded conservation table. Arrange the pieces with disciplined spacing and believable conservation handling. Include enough variation in scale, finish, and damage state to suggest excavation reality rather than a perfectly staged luxury still life. Nothing should look modern or decorative for its own sake; every object must feel like evidence tied to the same cultural world as the main topic.
Important details:
Emphasize tactile surfaces, honest aging, and historically grounded material behavior. Keep the architecture unmistakably Maya rather than generic fantasy ruins, with steep stepped geometry, roof-comb silhouettes, believable glyph blocks, and dense tropical vegetation that respects the stone volumes instead of swallowing them completely. Show abrasion, dust, corrosion, cracks, mineral bloom, pigment retention, or edge polish where appropriate. Keep the arrangement orderly but not sterile, as if prepared for a serious cataloging session. No readable labels, no museum placards, and no visual shortcuts that turn the study into product photography detached from archaeology.
Use case:
Maya Temple Dawn Discovery single study 6/9.
Constraints:
No readable text, no watermark, no logos, preserve exact 1:1 aspect ratio, no modern vehicles, no sci-fi glow, no fantasy monsters, no glass skyscrapers, no luxury advertising gloss, no floating objects, no velvet jewelry styling, no unrelated tools.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。





