
原始文档
GPT Image 2 · en
Scene:
A square frame from the same hillside viewpoint and distance as the reference image, now set in blue hour just after sunset. The launch valley is dimmer, lanterns and sodium work lights glow softly through mist, and the rocket stands in the same position beside the same gantry, reflected in wet concrete and narrow irrigation channels. Low clouds hang over the hills, and the scene should feel hushed rather than dramatic.
Subject:
Show the same orbital rocket during cryogenic fueling. White vapor spills slowly around the lower tank and pad deck, the umbilical arm is connected in the same place, frost bands build on the body, and a few tiny technicians in rain capes move near tool carts and hoses. The facility must remain the same one established in the reference: same tower silhouette, same service buildings, same ridge lines, same drainage canal, same patched stair railings. The emotional twist is that a technically intense fueling operation reads like an evening village ritual, with lanterns, puddles, and patient preparation.
Important details:
Preserve identity continuity from image 01: same rocket diameter, nose shape, engine skirt, gantry spacing, camera angle, and valley layout. The style should remain hand-painted and tactile, with watercolor skies, soft edge control, and subtle cel shading, but the hardware cannot become vague. Use cool indigo, moss green, cream, and warm amber accents. Keep the fueling vapor dense and believable near the ground, not stylized into magical clouds.
Use case:
Ghibli Countergenre series study 2/8, testing whether a realistic fueling scene can retain reference identity inside a soft animated-film atmosphere.
Constraints:
No readable text, no watermark, no logos, preserve exact 1:1 aspect ratio, no second launch vehicle, no fireworks, no military spectacle, no fantasy creatures, no cockpit close-up, no hard-edged neon cyberpunk lighting.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。





