Coloring a New Dimension

Siggy Ferstl at the Da Vinci Resolve system

About a year before Colorist Stefan Sonnenfeld here at Company 3 performed the color-grading work on Disney's 3D feature G-Force, a group of us under him started preparing for the industry's movement toward 3D production. The specifications of a 3D image as it's projected in a theater are very different from those of a 2D image, and that goes far beyond the obvious fact that there are two images (the left and right eye) being put on a screen simultaneously. The projector brightness is considerably less; the screen reflectivity, at least in the RealD process we used for G-Force, is considerably more; and the glasses people wear in the theaters have a tint to them. Those factors alone considerably alter the way color and contrast must be treated between 2D and 3D presentations.

Then there is the all-important issue in the 3D world of convergence. One of the biggest complaints audiences have had about previous (photochemical) versions of 3D have been the dizziness, headaches, and stomachaches that can result from technically imperfect and aesthetically exaggerated uses of convergence.

Preparing a Company 3 3D DI theater to be able to color our first 3D project was a major undertaking. Devin Sterling, executive producer of features and DI at the Arizona Avenue branch of Company 3, worked closely with Chief Engineers Mike Chiado and Jay Bodnar in the development of our 3D workflow. This involved installing a silver screen for working in the RealD format and ensuring the projection, software, and hardware were capable of handling the massive amounts of data involved in 3D DI work. They also specially adapted the light output of the Barco 2K projector to the RealD spec.

We also spoke with the people at Da Vinci Systems, of course—the company makes all of our color-grading consoles—about building a 3D system that could deliver two (or more) 2K image streams in realtime. And any 3D system from Da Vinci would need to allow subtle adjustments to the convergence during the DI sessions. Fortunately, Company 3 has the power of our enormous storage area network (SAN) so we didn't have any issues about pushing such a massive amount data through the system in realtime.

On the subject of the SAN, Company 3's Executive Producer, DI Operations, Devin Sterling says, "In all, G-Force took up about 30TB in its finished state. This was on our SANs—based on SGI CXFS DDN storage, which allows high throughput and realtime speeds directly off the SAN with speeds between 800MBps [and] 1.2GBps. Even with this kind of performance, we were not able to drag four streams of 2K data—right and left eyes and then the mattes we built for the right and left eyes—but it did give us solid performance for realtime stereoscopic playback of 2K streams and even some additional bandwidth to run other HD or 2K work without taking performance hits."

We were fortunate in that our first foray into 3D, G-Force, was originated as a 2D show. It allowed Sonnenfeld to grade the 2D version on a Da Vinci Resolve first and get it to a point where everybody was happy with that. This better enabled us to identify and separate out the issues unique to 3D grading.

So, you might ask, what would happen if we just applied his 2D grade to the 3D version of the film? Well, first of all, because of the nature of today's 3D spec (with one projector displaying two images at once) the overall screen brightness in a 3D show is considerably less than it is with a 2D show. This means that the definition of "white" on the screen in a 3D show is actually much duller (in an objective sense) than it is in a 2D show, and this affects the look of everything else in the image. Apply the 2D grade to a 3D presentation, and everything will look dull.

It was possible for me to build a system to translate the look of the 2D grade to 3D images through the use of look-up tables (LUTs) that do some of the number-crunching involved in getting the 3D images to share the visual attributes of the 2D grade. I built these LUTs inside the Da Vinci Resolve using its internal LUT builder. One of the difficulties in trying to get brighter whites was in doing so without clipping the image and losing highlight detail altogether.

While the LUTs enabled us to get about 80 percent of the way for the 3D version of G-Force, we still needed to go through by eye and make decisions scene-by-scene using Sonnenfeld's 2D version of the film as a visual reference. Some of the issues involved in giving the 3D version the same look as the 2D grade simply could not be quantified in the form of a LUT. After I built the LUT and did a pass of the 3D version of the film, Sonnenfeld went through with the clients to fine-tune that and come up with the final look of the film.

Related Links

It Came from the Third Dimension By the time Jeffrey Katzenberg announced last year that DreamWorks Animation would forevermore author all of its computer-animated feature films in a native 3D format (dubbed InTru 3D and developed in partnership with Intel), Monsters vs. Aliens was well into development as a 2D project... Digital Puppeteers Stop-motion animation and stereoscopy aren't techniques that audiences have ever seen combined before that is, until Coraline, the latest feature from... Stereo Hype This past December, Vic Love was headed to Jamaica to shoot footage for a documentary about the history of reggae music. With him traveled a tech-averse...

Convergence also required fine-tuning in the 3D grading sessions. While the visual-effects team had built convergence decisions into every shot, it wasn't until the entire 3D version of the film was put together and shown in Company 3's 3D DI suite that the creative contributors could really get a feel for the flow of the 3D effects in the film from start to finish.

This is a very important step in finishing a 3D film. There are certain triggers in a 3D movie experience that can cause the kind of eye fatigue and physical distress that helped kill previous 3D movements for audiences. Cutting from an image without much apparent depth to one with extreme apparent depth can cause eyestrain, as can forcing the audience to adjust to extremes of depth for protracted periods of time. Here it was very important that our Da Vinci 3D system enable us to fine-tune the convergence shot by shot. Again, I made some of these adjustments in my 3D pass, and Sonnenfeld fine-tuned these choices with the clients to set the final convergence decisions.

For certain sections of the 3D version of G-Force, we actually had to push the Resolve 3D even further than we expected because the filmmakers designed the 3D version to allow characters to pop out from the constraints of the 2:40 aspect ratio. This was done by designing the frame to be projected 1.85:1 and then containing the action inside a 2.40:1 matte. Occasionally, they would use the matted area to put action that seemed to jump off the screen.

In order to achieve this effect, we were supplied special mattes for the top and bottom of the 1.85:1 frame that could change shape to allow whatever action we needed to see in that area. To make this happen, we ended up having to send four imaging streams through our Resolve 3D: left eye, right eye, left eye matte, and right eye matte. Fortunately, the Resolve 3D and our SAN were up to the task.

Today's technology has moved us to the point where viewing 3D does not have to give audiences a headache or a stomachache. And we've gotten to the point at Company 3 where we can handle the technical requirements of grading for 3D, 2D, or both. Ultimately, the future of the format will have to do with how people shoot their films—what lenses they use, how fast they cut, and where the cuts are. Those will be the things that make audiences either want to see 3D movies or not.

Siggy Ferstl is a two-time master colorist of the year whose feature-film credits include House of Flying Daggers and Live Free or Die Hard. He joined Company 3, Santa Monica, Calif., part of Ascent Media, in May 2009.