I assume the video frames are each from a different reaction, or many many reactions. Then they are combined to create a model of a single reaction.