Reviews agree that this paper provides a technically elegant improvement to the state-of-the-art in video interpolation, an important problem with a wide range of practical applications. Results are qualitatively extremely impressive, in the supplemental video the video interpolation quality improvements over previous work is dramatic; this result should be of wide interest. Among its technical innovations, the paper introduces a novel "flow reversal" layer needed for video interpolation, and show how this differentiable mapping can be integrated in the deep motion prediction network.