This repository provides a PyTorch implementation of a neural network method for video-to-video translation at high resolutions (e.g., 2048x1024). The model translates semantic label maps into photorealistic videos, synthesizes talking heads from edge maps, and generates human motions from pose inputs. It extends image-to-image translation techniques to temporal video synthesis using deep convolutional networks.