Short Notes from "Cave of Forgotten Dreams" Post-production

by Kaspar Kallas

This 3D project by Werner Herzog and DoP Peter Zeitlinger, started for me by a phone call in the late evening offering me a DIT job on SI2K shoot in south of France. I had 10 minutes to decide. The rest is a documentary film…

 

On set Iridas tools were irreplaceable because we lacked proper tools to make this technical very demanding job on the first week of shooting. So there was quite a lot of duct tape and spit solution used to acquire the footage needed. This in turn required good tools for QC, to be certain that we would have something that we can use in a film.

The shooting period was split into two weeks of shooting with week and a half in-between. I used the in-between time to make a very solid side-by-side rig for use inside of the cave. There could be no adjustment screws as they would have become loose when the crew was climbing in and out of the cave. Because of Iridas native support for cineform RAW we could watch the dailies instantly when we left the cave to make necessary adjustments to our gear, workflow or both.

 

On the second week we had 2 sets of cameras, one would reside in the mirror rig while other were used for side by side. So when we needed to change in the interview we would just switch the heads to another setup while keeping the same recorder.

 

The production was extremely fast paced, up to 2h of footage daily. The postproduction was light-speed. I had exactly 2 weeks from the day I got the locked cut to the premiere in Toronto International Film Festival.

About 12minutes of the final 90minute film is non-synced go-pro cameras. I already knew the problem from set, as part of it was shot with one camera upside down the 3D was unwatchable because of rolling shutter tearing the images apart. There really are only 3 possible solutions to the problem today. Try to render a depth map based on original stereographic setup - extremely complicated because of the rolling shutter, compression and lens distortion artifacts. Use bezier shapes to push in or pull out parts of the image. Just use plain 2D slightly pushed into the screen.

 

The last option is worst, because it really pronounces all the problems with 3D cinema and leaves one wonder why to wear those geeky glasses in the first place. The first option was too time consuming and results were not really predictable.

2 days before screening, after 1st pass render I notice a bug, all the frames are not the correct frame size - a bug in beta version. Iridas was very prompt to deal with the problem in less than 12h I have a new build and everything is on the way again. The final version for the screening in the theater is taken to Toronto by Director himself just hours before screening as we hit another bug, this time in EasyDCP software that was also cleared within hours.

After 2 weeks almost no sleep, the show starts, I can breathe easily - the film is there, everything seems to look correct. 10 minutes before the end of the film the image goes black, sound continues for a moment and then it is stopped. Silence.

It turned out that it was the very first public screening in the Bell Lightbox for digital 3D and somebody had accidentally turned off the the AC for the projector and bulb had overheated.

The golden rule of filmmaking: if something can go wrong, it will go wrong, alligators or no alligators!