[片段参考] + 将参考片段视为摄像机从未停止录制。在其之前立即生成缺失的几秒，使用完全相同的镜头语言、光线、节奏、摄像机运动和连续性…

Here's the base prompt I use for every "Reverse Extension": [clip reference] + Treat the reference clip as if the camera never stopped recording. Generate the missing seconds immediately before it with identical lens language, lighting, pacing, camera movement and continuity, ending on a frame that seamlessly connects to the first frame of the reference clip. Build a natural cause-and-effect sequence that logically explains why the reference begins exactly as it does. Preserve the same characters, wardrobe, nvironment, camera language and visual continuity throughout. The final frame must visually match the first frame of the reference clip as if it were one uninterrupted take + [prompt scene ] This is the foundation. You can either let Seedance decide what happened before… or completely direct the missing sequence yourself.

[片段参考] + 将参考片段视为摄像机从未停止录制。在其之前立即生成缺失的几秒，使用完全相同的镜头语言、光线、节奏、摄像机运动和连续性，结束在一个与参考片段第一帧无缝衔接的画面上。构建一个自然的因果序列，逻辑性地解释为什么参考片段恰好以那种方式开始。 在整个过程中保持相同的角色、服装、环境、摄像机语言和视觉连续性。 最终画面必须在视觉上与参考片段的第一帧匹配，仿佛是一个不间断的镜头 + [提示场景] 这是基础。 你可以让 Seedance 自己决定之前发生了什么…… 或者完全由你来导演这段缺失的序列。