View-D: a revolution in stereoscopic 3D?

[NewsMaker]

Prime Focus's View-D process promises to convert 2D footage to stereo in record time. But can it really work? 3D World investigates


Image courtesy of Prime Focus

For several years, stereoscopic 3D has been one of the most active areas of R&D for post-production work. However, Avatar’s unprecedented box-office success has prompted a renewed surge of interest in the topic, with many distributors keen to explore the conversion of films not originally shot in stereo.

Released this month, Louis Leterrier’s remake of the Ray Harryhausen classic Clash of the Titans marks the first public outing of Prime Focus VFX’s proprietary 2D-to-stereo conversion process under its brand name, ‘View-D’.

In contrast to traditional frame-by-frame painting techniques, the system promises fast, semi-automated conversion. 3D World spoke to company president and senior VFX supervisor Chris Bond to find out how View-D works.


3D World: Let’s start with some context. Does the View-D process come out of your experience on projects like Journey to the Center of the Earth?

Chris Bond: Actually, I had worked on a 3D project way back in 1998 in Manitoba, and Frantic Films had [later acquired by Prime Focus] worked in VFX for almost a decade [so] we had worked on all sorts of projects where we were calculating 3D space and 3D volume. Out of that we had the basics for a 3D conversion pipeline, but it just seemed like time-intensive, difficult, slow-going work.

Traditionally, there are a couple of approaches [to conversion]: you can cut things out, move them around on cards in space and then paint out the holes frame by frame; or you can rebuild the scene in 3D, reproject onto geometry, and repaint textures you have to for every frame in the movie. Both those techniques are incredibly time-consuming.

My epiphany was the idea that we’d been doing things in a far more complicated way than we had to. Working with the R&D team, I spent evenings and weekends developing this concept, and came up with a way to do it almost interactively.


3DW: So how does the View-D process work?

CB: The View-D process basically breaks down into two stages. The first is selection. We use a variety of approaches to select objects we want to give depth to. Of course, you want to give depth to the entire scene, but how [the individual elements] separate gets set up at this stage.

And then in the second stage, our depth-generation tool pulls things forward or back in visual space, so [they] don’t look like they’re on cards; it gives them volume, dimensionality and depth. That tool is the ‘secret sauce’ where the 3D comes in. Because the process is fast, it almost becomes interactive for the artist.


3DW: Is the first step entirely automated, or does it require the artist to manually identify foreground and background elements?

CB: It depends completely on the content. We’ve written a number of object analysis and specialised keying tools, but if you have a crowd shot with hundreds of people crossing each other, you can’t necessarily run an algorithm that’ll detect all of those layers. In the best case, you’ve got automatic or semi-automatic solutions and in the worst case, it’s human effort – obviously computer-assisted, but it’s going to take time.

Sometimes, it will take a week before we can show the client the first frame of [the converted footage]. But there have been cases when we’ve had the client give us material at 7pm and come back at 9am to see it converted.


3DW: How does that client approval process work?

CB: When we start on a project, we pick ‘keystones’ from the film. [50–100 shots that form a representative sample of the client’s needs for the stereoscopic conversion work.] We take those through the View-D process, show the client the result, and get their notes.

At the same time, we’re bringing in the movie online and getting it prepped through stage one of the process. [By the time that’s done] we have hundreds of shots to work on. That’s the whole point of this process: rather than spending hours on one shot, we’re pushing out a lot of content, doing a lot of iterations.


3DW: What distinguishes a shot that can be solved semi-automatically from one that requires a lot of artist input?

CB: There are two issues. One is subjective and creative. For example, if you’re in a helicopter looking down at a city, in reality you don’t have much stereo vision [because of the distance of the objects relative to the separation of your eyes]. If you create stereo depth in such a shot, it tends to miniaturise the objects, as if you were looking at a trainset on a tabletop.

Clients can want to go either way. If you have a large-scale VFX shot, you might not want to give it a lot of depth, because of that miniaturisation effect. By the same token, some clients will purposely miniaturise things: for example, so a wide shot of the earth is not just flat. We have to accommodate both.

In terms of technical problems of a shot, the biggest thing is super-high-frequency detail, like bare branches of a tree in the foreground. The easiest thing for us is dialogue sequences, because we have a really good approach for making humans look good in 3D.


3DW: What about things that would pose a problem in a conventional stereo pipeline, such as volumetrics and dust?

CB: Actually, they help us, because those shots have a lot of dimensionality already. This is the interesting thing about our technique: it seems to work best on shots that are composed and shot pleasingly. When you look at projects like that, everybody says, “Oh God, it’s got all this aerial density in the smoke, and it’s got focus issues: that’s got to be really hard,” but they almost seem to pop into place. I wouldn’t say it’s easy, but they don’t seem to pose the challenge some other shots do.


3DW: On VFX shots, do you require any 3D assets from the effects houses?

CB: Even with a movie that was shot two or three years ago, getting original assets would be incredibly difficult, so the process was designed with the idea that we would never get anything.

There are cases where getting an asset from the VFX vendor will be immensely helpful – selection stuff like outlines or alpha elements – but it’s really about getting to final quality faster, not about making the result better. By the time they’ve dearchived the material, we might have done the shot anyway. We’re not dealing with a process that takes months: we’re dealing with one that takes days or weeks.


3DW: Some people have compared post-conversion to recolourising a black-and-white movie. Isn’t it going to be better to shoot in stereo from the start?

First of all, I should make it clear that I’m not against shooting in stereo. To me, View-D is just an alternative, in the same way that a director may want to shoot a film in anamorphic; then shoot some plates in IMAX for the VFX work.

But shooting in stereo creates its own visual artefacts: vertical alignment, left/right eye discrepancies, reflections that are not in the plane of the object, and so on. In a film like Avatar all of those things get removed or fixed in a post process, which is time-consuming and expensive.

I’m very confident we can get similar results. I believe that when our work is screened, there are going to be so many people who aren’t going to realise we didn’t shoot it in stereo!


Discover how the other VFX work on Clash of the Titans was created in issue 129 of 3D World, including the epic creature effects

Click here to buy issue 129 online

Click here to visit Prime Focus online



More...