We designed a spatiotemporal generative adversarial network which given some initial data and random noise, generates a consecutive sequence of spa-tiotemporal samples that have a logical relationship. We build spatial discriminators and temporal discriminators to distinguish whether the samples generated by the generator meet the requirements for time and space coherence. The model is trained on the skeletal dataset and the Caltrans Performance Measurement System District 7 dataset. In contrast to traditional Generative adversarial networks (GANs), the proposed spatiotemporal GAN can generate logically coherent samples with the corresponding spatial and temporal features while avoiding mode collapse. In addition, we show that our model can generate different styles of spatiotemporal samples given different random noise inputs. This model will extend the potential range of applications of GANs to areas such as traffic information simulations and multiagent adversarial simulations.