Don't know their exact setup, but it's a "chase cam" arrangement, where the camera is suspended on a line from the canopy. Various aerodynamic devices are used to keep the camera oriented forwards (plastic bottles, fins, windsocks shuttlecocks, etc), but most of these don't seem capable of steadying the camera perfectly. Various pilots are now mounting their camera on gyroscopically-controlled 3-axis gimbal sticks like this: http://www.gearbest.com/multi-rotor-parts/pp_176948.html?currency=GBP&gclid=CjwKEAjwnf2wBRCf3sOp6oTtnjYSJAANOfheRwgcpyVWiRAIyQ30V6Gg3A2VFhSaqy3bRjVx35-GRBoCbdDw_wcB
Or similar 2-axis stabilizers designed for drones (that are often very competitively priced) like this: http://www.banggood.com/FPV-2-Axis-Brushless-Gimbal-With-Controller-For-DJI-Phantom-GoPro-3-p-908068.html?p=MW2604548780201404KC
Getting stable video makes a fantastic difference to the viewing experience. The new generation of gimbal chase-cam videos are really amazing, and give a steady, clear and unique perspective of paragliding. But it's not without problems and can be expensive, risky and hard to set up. I'm still trying to develop an aerodynamically-stable platform for my cheapo dangle-cam. Complete stability might not be possible but eliminating as much movement as possible is helpful - even when using a gimbal - because gimbals only auto-stabilize within a limited range. Gimbals are quite expensive, chunky, hazardous, need setting up and consume considerable battery power.
It looks like the Goldsmiths are using a 2-axis gimbal and leaving yaw to the aerodynamic stabilizer. Pitch and roll seem to be effectively stabilized but yaw (left to right movement) is still a bit uneven. Towards the end of the video you can see the boom and tail feathers they use to aerodynamically orientate their camera. I have tried similar devices and think that this design lacks drag so swings back and forth and doesn't dampen yaw enough... and explains why left to right and back and forth motion in this video is still quite variable.