Lockstep is a video model that replaces the people in a source video with characters from provided photos. It preserves the original movements, lip synchronization, and camera angles.