This paper presents a novel spatiotemporal action editing framework that synthesizes spatiotemporally synchronized multi-party interaction 3D video scenes from separately captured data. Our main idea is to model multi-party interaction events with synchronization among body actions and gaze actions from multiple objects, and use them as constraints to perform spatiotemporal editing on 3D video data. To achieve that, we first propose the idea of Action History Volume to enable to model multi-party interaction events into spatiotemporal constraints. Next, we propose a novel 3D gaze estimation method, taking advantage of the symmetry prior and the super-resolution technique. Finally, we propose a three-step action editing processing scheme, including data segmentation, intra-key segment editing and inter-key segment optimization. Experiment results prove the effectiveness of the proposed method.