ON SHOOTING CAVE OF FORGOTTEN DREAMS
by Peter Zeitlinger
We shot Cave of Forgotten Dreams on SI 2K cameras, GoPros and two tiny Canon amateur cameras.
We decided to shoot in 3D only a few weeks before we began production. My contacts at Pille Film and P+S Technik in Germany (who developed a system to handle SI 2K cameras for Slumdog Millionaire) were very useful for finding the state of the art in high-end small camera technology. Their publications all over the world promised to provide us with all the tools we would need in the cave, such as light weight Steadycam 3D rig by P+S and a small recorder for the SI2K streams by Pille. Unfortunately, both companies left us shortly before the shoot because they felt they were “not ready to guarantee us 100% of function of their prototypes.” Even I told them we had no choice: we don´t need a 100% guarantee, the Germans don’t do half things. So we had to go it alone.
Dave Harding, the unit production manager, found David Blackham in the UK who had some SI2k Cameras and a connection to British Technical Films. They had experimented with a Phantom Camera mirror rig and used it successfully several times in commercials. The computer hard- and software support came directly from Ari Presler, who is the father of the 2SI2K software.
The first time we entered the cave we had to shoot right away. There was no scouting. Werner Herzog was the only one from the crew who has seen the cave a few months prior. We carried the clumsy mirror rig with us provided by British Technical Films. It was used on several commercials under studio conditions before. After only a few meters inside the cave we decided to leave it behind us, because it was not possible to squeeze it into the narrow tunnel.
Since we had only a few hours to shoot the whole film we had to shoot no matter what. Werner said, "Take a gaffer tape and glue or something. Put the cameras side by side and let’s go." “I had been able to build a macro-extension out of a toilet paper roll in an Antarctic tent for filming inside a scientific microscope,” I answered, “but we should go back and try it the next day.”
After a moment Werner handed me two magic arms camera mounts. "Can’t you use this?" I ripped the cameras out of the mirror rig and fixed them side by side on the magic arms. 10 minutes later we started to shoot the secret paintings of the cave. I shot from the hip without a viewfinder. We skipped the complicated alignment of the 3D cameras and fixed it in the evening in Cineform (Software).
During the night I experimented with a horizontal configuration of the two cameras within the mirror rig to make the arrangement smaller. After a few experiments I dropped the idea. It was too wonky with the means we had at hand.
We found a local smith and he milled an aluminum bar with several holes and slots so we could mount the cameras on the same horizontal level side by side. The angle and the intraocular distance could be adjusted easily and quickly by loosening one of the screws before every shot. I was flexible for every distance and focal length and quickly got the feeling for the right adjustment without any complicated formulas and calculations.
The amount of people in the cave was limited because of a lack of oxygen and a high level of carbon di- and monoxide. Our stereographer was not admitted to the cave and he was busy with the skybot aerial rig anyway. Chris Watts, who controlled the skybot, shot marvelous aerials from the locations around the cave.
The computers needed an expert to be operated, and to make slight adjustments on the software. This was the moment when we invited Kaspar Kallas from Estonia, an expert in digital matters, 3D software and Hardware issues. I called him “Tuttle” (the plumber from the film Brazil).
On the next day we mounted two GoPro Cameras on the new bar. There was no frame lock possible on those cameras. This was also an issue with longer takes. The footage of the two cameras came apart after a few minutes. We did an optical slate at the beginning and the end of a take so we knew how much of the running time the cameras were wrong and pulled the images together—up to one minute was not a big issue.
To get the lenses closer together, I turned one of the cameras upside down. Later we found out that the cameras worked with a rolling shutter that caused deformed objects in pans and other movements. And the deformation was different in the left and right eye. Kaspar Kallas tried to solve that problem by working on a software solution but at the end his company in Estonia has dealt with the moving objects manually in every frame. They used basically the left eye and added only a few important elements from the right eye to complete the 3D image. Using hand painted mats they spent weeks to pick out the image parts and fit and align them for the 3D vision.
The easy part of our 3D film was the interviews. We used the mirror rig from British Technical Films. Here we had only to deal with the different coatings of the lenses and the color shift created by the mirror.
Setting up the correct configuration took too much time and Werner had to be very patient watching us screwing around, aligning the cameras on the 3D test chart and adjusting the C-motion that controlled the stop and focus electronically.
Recording the stereo data was not easy either. The CPU of our computer tended to overheat and we had to restart it several times. Kaspar Kallas brought his own hand-made computer and bought a big fan from a local dealer and attached it with gaffer tape to the housing of the computer.
He did the same with the field recorder that recorded a double stream of data from the synchronized SI2K cameras. He built a switch into the fan so we could turn it off while shooting. I used the field recorder as a viewfinder—it was basically a tiny powerful computer. I synchronized the focus and the stop for the cameras with a belt built out of gaffer tape.
For the second half of the shoot he also built a mechanic gear to run the focus of the super 16mm Zeiss lenses simultaneously, and British Technical Films built a lighter mirror rig for the second part of the shoot.