X2Face: A network for controlling face generation by using images, audio, and pose codes

The objective of this paper is a neural network model that controls the pose and expression of a given face, using another face or modality (e.g. audio). This model can then be used for lightweight, sophisticated video and image editing.