原始文档
GPT Image 2 · en
Scene:
A square material-focused image where lighting, scale, and surface behavior are the main event. Build the frame around the dominant material language shown with full confidence. Keep the lighting, camera distance, and spatial organization tuned for a high-signal research image rather than a casual snapshot. The image belongs to the Complex Optics Materials theme and should feel immediately legible as a prompt atlas theme for complex optics materials, testing photorealism, physics-materials, reasoning-composition with gpt image 2..
Subject:
Center the study on the dominant material language shown with full confidence. Translate the theme into a concrete visual statement instead of an abstract mood board: the viewer should understand what kind of image this is, what is being tested, and why the composition is difficult. Use the slug idea of complex optics materials as a literal guide for subject matter, materials, and staging.
Important details:
Microstructure matters: refraction, fibers, scratches, pores, edge wear, translucency, and dust should all feel physically specific. Preserve a disciplined square composition, clear focal hierarchy, and enough secondary information to make the image feel complete rather than underspecified. The image should look polished, novel, and exhibition-ready, with topic-specific credibility coming from the way surfaces, structure, and visual language reinforce Complex Optics Materials. Avoid generic filler elements that do not serve the main concept.
Use case:
Complex Optics Materials single study 1/8, testing photorealism, physics-materials, reasoning-composition through the surface hero variation.
Constraints:
No readable text, no watermark, no logos, preserve exact 1:1 aspect ratio, no generic stock macro, no broken scale cues, no plastic-looking texture when it should feel natural, no meaningless blur, no watermark, no fake labels.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。