“Google's Omni model can both input and output video, and allows users to interact with and edit the video through natural language.”