Lip Sync Videos

According to Microsoft's official press release, Microsoft today announced a VASA-1 framework for graph-generated video, an AI framework that uses only a portrait photo of a real person and a piece of audio of a personal voice to generate accurately lifelike lip-synced video (video that generates a reading of a script), which is claimed to be particularly natural in terms of expressions and head movements. Much of the research in this area has focused on lip-syncing, while facial dynamic behavior and head movements are often overlooked, resulting in stiff, unconvincing faces and the Valley of Terror. Microsoft's VASA-1 framework overcomes the problems of previous facial generation techniques...