Talking Face Generation via Learning Semantic and Temporal Synchronous Landmarks | IEEE Conference Publication | IEEE Xplore