The quicktime shows a playblast of the particles making up the bubble simulation. The behavior is a result of Fusion's proprietary bubble model, which realistically simulates the forces around the outside of the bubbles and results in extremely realistic toroidal flow on the inside of each bubble. The model reproduces all the characteristic motions of air bubbles rising in water, unlike the usual simplistic approach of just having jiggling balls animated to rise upward. In our model, we get the characteristic arcuate shapes of the bubbles (vertically squashed by buoancy forces), side-to-side oscillations in their path as they rise, bubble-size-dependent motion (faster rise for bigger bubbles, more oscillations), dynamically deforming shapes, turbulent wakes that rip off pieces of the bubbles, and coalescence of bubbles when they make contact with each other. They also react very realistically to solid collision geometry.
The bubble behavior was created using our proprietary physical model (python-scripted). It is possible to get a highly simplified version of the bubble behavior by just using an upward gravity field, some noise to jiggle the bubbles as they rise, and strong surface tension. Obviously you won't get all the detailed extras that cross the boundary between cg and photoreal -- such as the vertically squashed len-shaped forms, the lee-side ripping, bubble-size-dependent behavior and shapes, and side-to-side oscillations. For that you have to create your own physical model. The physics for this is quite well known, at least in terms of empirical relationships, it's just a matter of adapting it to work with a SPH solver.