Despite the recent success of single image-based 3D human pose and shape\nestimation methods, recovering temporally consistent and smooth 3D human motion\nfrom a video is still challenging. Several video-based methods have been\nproposed; however, they fail to resolve the single image-based methods'\ntemporal inconsistency issue due to a strong dependency on a static feature of\nthe current frame. In this regard, we present a temporally consistent mesh\nrecovery system (TCMR). It effectively focuses on the past and future frames'\ntemporal information without being dominated by the current static feature. Our\nTCMR significantly outperforms previous video-based methods in temporal\nconsistency with better per-frame 3D pose and shape accuracy. We also release\nthe codes. For the demo video, see https://youtu.be/WB3nTnSQDII. For the codes,\nsee https://github.com/hongsukchoi/TCMR_RELEASE.\n