Create a cinematic 15-second vertical 9:16 surreal urban short film, photorealistic with bold graphic shadows and one coherent visual sequence. 0–2 seconds — VISUAL HOOK: At night, pedestrians wait at a city crosswalk under a bright red traffic signal. The instant the light turns red, every human becomes completely frozen in place—but their shadows continue moving independently across the pavement. One shadow looks upward, then urgently signals the others to follow. 2–7 seconds — ESCALATION: The camera begins at street level and tracks sideways as the living shadows detach from their owners. They pull together black road markings, building shadows and pools of darkness, folding them into a physical staircase made entirely from shadow. The frozen humans remain motionless while only the shadows work rapidly. 7–11 seconds — CAMERA REVEAL: The camera cranes smoothly upward into a high top-down view. The staircase grows vertically along the side of a building, then continues impossibly through the sky toward an enormous full moon. The shadows run upward together, their flat silhouettes stretching across the steps. 11–15 seconds — PAYOFF: The traffic light turns green and the humans resume walking as if nothing happened—but none of them cast shadows anymore. The camera tilts from the shadowless pedestrians toward the moon. Tiny silhouettes of the escaped shadows celebrate and dance across the moon’s surface. Hold the final moon image long enough for the twist to register. High-contrast wet streets, reflections, deep blue night, bright red and green traffic lights, readable silhouette animation, consistent pedestrian positions, controlled camera geography, mysterious yet playful mood. No dialogue, no text, no subtitles, no horror faces, no random transformations, no duplicated pedestrians. Sound design: distant traffic, electrical signal click, soft scraping shadows, rising atmospheric pulse, faint celebratory footsteps from the moon.
本片段由 ByteDance 的 Seedance 2.5 生成,页面上的提示词就是生成它的原始文本。
示例由 Seedance 2.5 生成,参数:427:240 · 0:15 · Surreal · Short Film · Cinematic。沿用这些参数可获得最接近的效果。
先原样复制提示词,只改主体、保留镜头/运动/光影描述。视频模型对措辞很敏感,多试几次为佳。
当然可以。提示词最适合作为起点——替换场景、调整节奏词或宽高比,适配你的项目。