You can download an MP4 version.
Technocrane - used to record the original camera moves
"Could Take4D playback pre-recorded camera moves on a gyro stabilised helicopter camera flying at 100mph 20 feet above the trees? At present, there is no system in the world that can do it - you only have 5 days - 2 of which are the weekend.".... Joe Lewis - General Lift.
In the feature film Hansel & Gretel - The Witch Hunters, there was a scene where a Witch was flying on broken tree branche with a character hanging off the end.
The actors were suspended on a wire flying rig and General Lift recorded the Technocrane camera movements using their EncodaCam system.
These action sequences were originally shot against green screen in Babelsberg Studios - Germany, where the green was replaced in real time on-set with an arbitrary CG rendering of a night time skyline of trees.
The green screen was to be finally keyed against helicopter background plates shot over the Bavarian Alps, Germany, and the VFX Supervisor, Jon Farhat, wanted the backgrounds to match the camera work done in the studio.
Richard Widgery of Take4D was approached to handle the development of a motion control system to achieve these shots.
Helicopter Film Services were already signed up for the job as Aerial Cinematographers and were going to use a Nettmann Super-G gyro stabilised rig with a RED Mysterium-X camera. Richard's task was to take the camera moves and find a way of playing them back on the Super-G.
The Super-G was originally designed as a self contained system with its own joysticks and control panel with no external interfaces at all. Nettmann provided Richard with their own proprietary communications protocol so that their control interface could be removed and replaced with Take4D running on a laptop.
The data from the Technocrane was essentially a text file containing position and rotation information of the camera head movement. The position information was obviously going to be ignored since there would be no way to control the helicopters flight path, but the pan, tilt, and roll information needed to be sent to the Super-G. Unfortunately the Super-G control system is open ended - no feedback at all - which means that there is no way to know where the camera is pointing at any given time. It is not possible to send it rotational commands such as "pan to 25 degrees to the left". The only signals it accepts are voltage commands, and it uses these to directly drive the motors. The longer the voltage is applied, the more the head rotates.
The EncodaCam data had to be converted in real time into voltage signals to send to the Super-G. Richard also had to develop a complex calibration system that could accommodate for drifting and non linearity of the rig. These had to be set at ground level and then tweaked whilst flying. Once up in the air, the temperature would change with altitude and throw the whole system into inaccuracy.
Much more practical issues were that of space and vibration. It was going to be cramped in the back of the helicopter. There would be three people in the back of the aircraft; Jeremy Braben (Aerial DOP) with his 15 inch Nettmann control system on his lap (for some other shots and zoom & focusing during the playback of the motion control data), Jon Farhat (VFX Supervisor / Aerial Director) with a 17 inch monitor on his lap, and Richard, with a 17" Take4D laptop for the motion control playback. Typing or using the "track pad" on the laptop was going to be extremely difficult. Richard decided touching the laptop had to be removed entirely except for selecting and loading up the next shot which could be done with 3 key pressed. Everything else had to be taken off the laptop onto a more convenient interface.
The solution he chose was an X-Box controller. The right thumb pad was used to pan and tilt the camera; the left was used for roll. The buttons were programmed with a variety of function to play the move forward or backwards, change the calibrations, compensate for drift, and emergency stop.
With the bulk of the development complete, the equipment was shipped out to Munich airport in Germany where it was fitted to MHS Aviation's Aérospatiale AS 350B3 Ecureuil Helicopter.
An added and rather unforeseen complication overshadowed the success of the whole project. Iceland's Grimsvotn volcano was erupting and its ash cloud had drifted over Germany (BBC's coverage). Any aircraft that used a Jet engine, which this helicopter had, was grounded. Fortunately, at about midday, the cloud which had descended half way south through Germany stopped and the retreated north, leaving Munich airport free to re-open.
This had however pushed the timetable back and as of lunchtime, no test flight had yet been done. Whilst everything worked fine on the ground, it was uncertain if the same would be true flying at 80 knots (~100mph). In-fact, the wind speed was having some effect on the data and Take4D was quickly modified, enabling the X-Box controller joysticks to manually compensate for this force whilst the move was being played back (a real time layering of motion control and manual control). This was only added in the cockpit just minutes before takeoff to shoot the final footage (see picture left).
With everything in-place, the team took off for a 6 hour round trip flight over the Bavarian Alps. The first half was for non motion control shots for which Jeremy used the normal Nettmann control system.
Flying at such low altitudes, small bugs colliding with the lens shield was a slight problem. Anything off centre, Jon was happy to paint out in post production, but head-on marks required a landing to clean the glass. Landing in the Alps was not the easiest of tasks and really showed off the skill of the pilot. Taking off without getting the dust and dirt blown up from the downdraft demonstrated just how powerful these aircraft really were, ascending faster than any debris.
During a refuelling stop in Innsbruck, Austria, the Nettmann pan / tilt / roll controls were removed and replaced with the Take4D control system for the return flight.
Once flown to the required location within the mountain range, and with the helicopter hovering straight and level, Richard used the X-Box controller to re-tune the systems calibration to eliminate any drift that may have been introduced by the difference in air temperature affecting the Nettmann amplifiers. Without this calibration, the camera would slowly turn in one direction or the other. With the drift eliminated, the camera was targeted head level and straight forward in order to "zero" the system. This reset all pan, tilt, and roll values relative to the helicopter in order to replicated the configuration of the original EncodaCam system used to record the data on the Technocrane.
Using just 3 key presses, the first camera move data was loaded into Take4D and converted ready for playback on the Super-G. This produced a "first positions" numerical read-out for the required pan, tilt, and roll for the first frame of footage. The X-Box joystick controller was then used to drive the camera to match these values.
The camera was now set in its first position ready for recording, and a signal was given to the pilot to start their run. Once the helicopter was up to speed, and on the command to shoot, the start button was pressed on the X-Box controller and the playback of the motion control data commenced. Whilst keeping an eye on the footage, Richard could use the X-Box joysticks to compensate for any wind effect that was causing the camera to be blown off course, whilst Jeremy was simultaneously controlling zoom, focus and iris using his usual Nettmann lens controls.
For each consecutive shot, the data was loaded with the system constantly live and the camera driven into the new starting position using the X-Box controller. Occasionally amplifier drift was re-corrected to keep the head stable again using the X-Box controller.
23 day for night shots were recorded in rapid succession as the sun set, requiring the team to get each of the shots first time, in a single take. Both the X-Box controller and minimal use of the laptop proved their worth as there was no margin for errors or delays.
With all the shots recorded and the sun gone, only a long wait remained to find out if the plates matched the original footage.
A few months later the call came through from Joe Lewis - "It worked out great and the plates dropped straight in".