Our data format is capable of generating unseen views from multi-view images. Thus, realistic fake videos, whose camera trajectories are different from the original video, can be generated from our dataset.