We present a deep learning based technique that enables novel-view videos of human performances to be synthesized from sparse multi-view captures. While performance capturing from a sparse set of videos has received ...