Not sure if this is how it's done, but you can achieve this effect if you take several frames from a shot where there is camera movement and morph in between them.
So let's say you pick the following frames from a second of footage (25 fps example):
You would morph from 1 to10 in 25 frames, then from 10 to 1, again from 1 to 10 and then 10 to 18. Then 18 to 24, 24 to 18 , 18 to 24, 24 to the next frame you pick up. You may need to play around a bit until figure out the frames you pick and the amount of frames in which the morph happens so that you can match the beat.