In this paper, we propose a method that seamlessly pastes a source video object into a target video using the 3D Poisson equation. The proposed method extends the 2D Poisson equation, which is used for the seamless image composition method, to the 3D equation in order to consider both spatial and temporal coherence. Furthermore, it divides a long video into a set of short image sequences and solves each problem separately, then merges them using linear blending in order to avoid a memory leak problem. As a result, the method can generate the temporally-coherent composite video, while the frame-by-frame composition using the 2D Poisson equation suffers from side effects such as a popping color effect.