Interview with Howard Postley, 3ality Digital Systems - Part 2

Studios seem to be very careful about their 3-D releases in order to not hurt or disappoint movie-goers. Could bad content slow down or even break the 3-D wave?
Without question. Anything that causes the audience to think that 3D in the 21st Century is the same as what was made in the 1950s is bad. If the audience decides that 3D is just a gimmick or a passing fad or hard to watch or just another effect or just being used to try to compensate for other weaknesses then it changes from a positive to a negative and the wave will die out.

You also have to keep in mind that studios are careful for financial reasons, as well. 3D theatrical releases may do 3x the per screen revenue of the same release in 2D but those numbers do not translate to the home audience yet and that is where the money is made.

What are the main mistakes 3-D content providers should avoid?
I don't want to tell people what they should or shouldn't do. I will tell you some of the mistakes that we have made, however. The number one mistake we have made, and continue to make occasionally, is forgetting that 3D is not 2 times 2D. This applies to almost everything; composition, acquisition, post, distribution, infrastructure, etc. Let me give you an example. Since the market for 3D is smaller than for 2D, most productions start out thinking of a 2D "product" with a 3D add-on. Some take the opposite view and think about making a 3D product that they can pull a 2D version out of. These ways of thinking would be fine if the difference between 2D and 3D was just equipment, but it isn't. There are a lot of creative differences, as well. That isn't to say if you want a 2D version and a 3D version of something that you have to shoot two separate movies but you do need to plan for both, not just assume that one is a subset of the other.

Another somewhat different example would be a live football game. Here is a case where you can pull the 2D from the 3D cameras, but you don't just replace all of the cameras at each position with a 3D camera. The 3D camera angles that you'll want are different from some of the 2D angles and, generally, you don't need as many 3D cameras. Also, the pacing of the 3D show is different so you'll need a separate Director.

There are so many things in 2D that are well understood and seem like they should translate easily to 3D, but they don't. Things that seem, on the surface, to be relatively simple often turn out to be complex.

Regarding 3-D shooting, there are a lot of diverging opinions regarding interaxial distance settings, parallel vs. toe-in, depth of field, etc.? What would be your recommendations for a good 3-D captation?
My opinions on these topics really isn't relevant. I am not a cinematographer. I'm not qualified to shine Peter Anderson's or Steve Schklair's or Max Penner's shoes on this issue. What I can say, from a technical perspective, is that a lot of the diversity of opinions on these topics are based on the legacy technology of the past, not that of the present. For instance, one of the main arguments in favor of parallel shooting is that technique is far less likely to create excessive background parallax. In the days where one had to get out a calculator and a tape measure (or a laser) and setup these shots, that might be a significant factor but the tools of modern 3D production are so much more advanced that this kind of issue shouldn't be a factor at all. Camera controllers can monitor and manage this type of situation to prevent the problems and allow the focus to be on creative choices.

Which features did you need to implement in your camera rigs in order to make them really usable on the field?
What makes our rigs usable in the field is that they are fast and easy to work with. We have automatic alignment of cameras and lenses, automatic matching of zoom and focus, automatic protection against excessive parallax, and automatic correction of color wedges, geometry, keystoning, etc. The entire alignment process takes a few minutes. The 3flex SIP stereoscopic image processor constantly measures dozens of parameters about the rig, as well as synthesizing metrics about the stereoscopic image, and alerts the operator to those about which he or she should be aware. Everything is actively managed by the automated systems. It isn't a situation where you reset everything for every shot; the system is constantly adjusting itself. Over the course of a long shoot, especially an outdoor sporting event, environmental characteristics, such as temperature, will cause massive changes to the cameras. It used to be that someone would have to re-align the cameras in the middle of an event but that isn't necessary anymore. As for the rigs, themselves, ours are carbon fibre to make them very rigid but light weight. They are made to tight manufacturing tolerances. The motors provide high-torque but deliver it smoothly for high positioning accuracy.

Do your rigs need dedicated and trained shooting teams or are they accessible to any 2D operators?
When we shot at the SuperBowl several years ago, all of the 3D cameras that we used were operated by NFL Films cameramen who had never seen the camera systems more than a day before the game. Each of those cameras did need a 3D operator, who did need to be trained in 3D, however. Now, we have the ability to have a single stereographer manage the 3D settings of all of the cameras in a multi-camera shoot. We have some events coming up where we will use as many as 12 cameras but the 3D settings will be handled by one guy and those settings are creative choices that we wouldn't want to automate.

What kind of metadata is produced by your rigs how is it used on live captations and post-production workflows?
There are two categories of metadata for us: the physical metadata from the cameras, the head and the rig, and synthesized metadata that is calculated by the image processor. Physical metadata includes things such as interaxial distance, convergence distance, zoom position, focus position, tilt angle, pan angle, etc. Synthesized metadata includes items like, depth range, average depth, center depth, focus match, zoom match, luma match, chroma angle match, etc. In total, we produce about 100 channels of metadata. All of that metadata is sampled or calculated several times per frame and is timecoded. In post, some qualitative metadata are used for things like dailies. Some positional metadata and depth maps are used for compositing and effects. Other metadata that characterize the shot are used for various compensations at presentation time.

What measurement and monitoring tools are you using to validate your 3-D shots on set?
Depends on the type of show it is and who is looking at it. We usually have a color critical monitor and a couple of 3D monitors, in addition to a variety of standard monitors displaying things such as disparity maps, qualitative analytics, etc. The 3flex SIP image processors have a suite of scopes in them, both the usual color scopes: waveform monitor, vectorscope, etc., except that they are for stereoscopic sources, as well as a variety of 3D geometry scopes. The qualitative metrics of the 3flex SIP tell us whether a 3D shot is valid and watchable on different screen sizes. It is the job of the stereographer to determine whether a shot is creatively valid.

How do you take into account the final screen size?
Typically we dial the ideal screen size into the 3flex SIP so that it will tell us about the shots relative to that screen size. This is one area where some of the metadata from the SIP are used to inform downstream devices how to handle different, generally smaller, screen sizes. Usually, when 3D is displayed on a smaller screen that for which it was intended, there are no issues with watchability. However, reducing screen size also reduces the amount of depth. It also tends to affect things like floating windows, maximum positive and negative parallax, parallax transition rates, etc. We have the ability to take those things into account and adjust them.

How to shoot when the same content is targeted to various screen sizes (theatres, home displays, mobile phones)?
In essence, this is sort of the same question as the previous one. The bottom line is that there isn't one screen size, even for theatrical release. I have to say that, right now, the mobile environment is quite challenging, both because of the small screen size and the relatively low resolution of those screens. 3D perception is dominated by the impact of parallax which is related to absolute distance, not pixel offsets. As you go down in screen size, you have to make more use of other depth cues, such as color, to recover the desired perception of depth.

Why traditional camera manufacturers didn't provide any 3-D equipment so far?
I think that, in terms of cameras, themselves, the market hasn't known quite what it wants yet. I think that the camera manufacturers will get to a point fairly soon where two cameras can be connected to each other and controlled as one camera. However, there are limits to that. You can't, for instance, simply paint two cameras together. The colors need to match, not the color parameters. Lenses, sensor positions, etc. all need to be balanced relative to one another. Could you make a single stereoscopic camera? Sure, there have been a number of them. Would such a camera deliver the range of capabilities that are needed? No one has delivered that yet. Certainly, one could make a camera with more resolution in one axis and move a lot of what is currently in a 3D rig into a lens or a lens adapter but I'm not sure that's really all that different from a camera rig today. You're going to have the same issues with lenses and positioning. You'd get rid of rotation issues but you'd lose significant positional flexibility.

Beyond cameras, this is a lot of new stuff. Frankly, new categories of things generally don't come from large, established companies with entrenched product lines, they come from small companies that are very close to the problem and adapt quickly. It isn't like a lot of the equipment that we use in 2D, beside the cameras, comes from traditional camera manufacturers anyway. From an equipment point of view, this is a very small industry.

What would be your recommendations regarding 3-D post-production issues (re-convergence, floating windows, color grading, etc.)?
I think that you're going back to that place of creative decisions that should be made by someone other than me. My caution in these areas is that there is far more disinformation about how these things should be done than there is correct information. Most of the information is opinion, not fact. We spent an inordinate amount of time on U2 3D "following the rules" of 3D before we realized that most of the rules were obsolete or wrong to begin with. I would also like to caution people that most of the tools out there for doing 3D post are fairly primitive. In the hands of an expert, they can achieve excellent results. For others, it isn't hard to make a single shot to cause so much eyestrain that the viewers won't be able to watch the rest of the movie. It can take only a few seconds of bad 3D before eye fatigue overcomes a viewer's ability to fuse images. It can then take tens of minutes before he or she can overcome that fatigue. Convergence adjustment is more than simply changing the separation between two images. Color provides depth cues in 3D so grading is vastly more complex than simply coloring one eye and then matching the other one two it.

The whole situation of 3D post sort of reminds me of the early days of desktop publishing. When first put together, a Macintosh, a LaserWriter and PageMaker occasionally delivered beautifully typeset pages. Far more often, what we got were high-tech ransom notes. In the case of DP, the new discipline got enough time to develop and mature.

Part 1 here