It’s “Tilt-shift photography”. (Well that and some time compression.) The gist of it is this: you tilt the camera lens respective to the sensor, meaning that the focal plane is no longer parallel to the sensor. This basically screws up the depth-of-field information and for some odd reason the human mind interprets that as “miniatures”. That footage is shot at Sydney Harbor and apparently doesn’t even have post-processing. If you want to see more the photographer is Keith Loutit and his blog has more goodness. Check it out.