Realistic Speech-Driven Talking Video Generation with Personalized Pose

Figure 3

Dilated Depthwise Separable Residual (DDSR) unit network.