NeuMan reconstructs both the background scene and animatable human subjects from a single video using neural radiance fields. The method enables novel view synthesis and novel pose animation of the reconstructed human. It builds on PyTorch3D and uses deep learning for 3D reconstruction and rendering.