Temporal continuity requires a video model to preserve relevant state while time advances. Faces, clothing, object positions, scene layout, lighting, and camera motion should change only when the action or edit justifies the change.
Complex movement makes continuity harder because more relationships must remain stable across frames. Evaluation should inspect full clips and transitions, since individual frames can look convincing while identity or geometry drifts over time.
ELI5
Temporal continuity means a video keeps its world coherent as time passes. People, objects, backgrounds, and lighting should not change unexpectedly between nearby frames.
For example, a cup placed on a table should remain the same cup in the same place until someone moves it. A clip can contain beautiful individual frames but still feel broken if the cup or background keeps changing.
