Just a few years ago, after films like Avatar and Alice and Wonderland had helped spark the scramble to make 3-D movies, there were as many as 20 to 30 companies vying to offer the tools that let studios convert their movies to the new format. Among them was a Vancouver firm called Gener8, headed up by Rory Armes, a veteran of the video game industry who had a novel approach to the 3-D puzzle.
“3-D at that point was a really interesting wild west,” Mr. Armes says. “Everyone was out there mining.”
But today, that number of big players in the 3D space has shrunk to about four, and Gener8 is one of them.
A startup in 2010, the firm now employees hundreds of 100 developers and artists. It’s worked on films like Prometheus and The Amazing Spider-Man, and is currently taking the leap into the full-blown blockbuster scene.
In fact, Gener8’s approach was so novel, that film players wondered if it could possibly work. All 3-D movies achieve a sense of depth by sending slightly different images to viewers’ left and right eyes, producing a stereoscopic effect. As Tim Bennison, Gener8’s COO and lead technology guy, explains it, there are two ways to generate these parallel images. The first is to shoot a film in “native” 3-D, using two cameras at all times to generate the left and right-eyed pictures. But this setup can be very fiddly for directors – it’s bulky, slow to work with, and if the cameras get out of alignment, the results can be difficult to fix after the fact.
That’s why many films actually add the 3-D effect after the fact. Digital artists go through the film with software tools, manually selecting which areas of the frame should pop out and by how much. But this approach has its drawbacks as well: Human artists lack mathematical precision, and relying on artistic judgment from shot to shot can lead to inconsistencies, subtle or glaring, in how much any given element pops out.
“You’re not necessarily conscious of it, but your visual system will go, there’s something weird here,” says Mr. Bennison.
Gener8’s solution is something the company calls “virtual native” conversion, using video game technology to achieve the best of both worlds. Their software generates a 3-D model of the set, the actors and the action, and then uses that model as a reference for how much different parts of the frame should pop out for the viewer. The system still uses human artists to complete the job, but layers their work on top of the precise 3-D model.
“We can literally re-film the scene from the two virtual cameras,” says Mr. Bennison. “We can generate a left and right eye [perspective] which is precisely what you would have gotten if you’d filmed the actors.”
It’s an approach that tracks closely with the way visual effects companies model scenes before adding the requisite explosions and dragons. In fact, Gener8 uses industry-standard file formats and can pass its work files back and forth with effects houses as they build scenes.
The company was initially met with skepticism when it entered the Hollywood arena. But the firm got a toehold in part by doing conversion work on films like Prometheus, which was mostly filmed in native 3-D, but needed some non-3-D scenes converted after the fact. Mr. Armes says Gener8’s results were indistinguishable from the real thing.
The attitude, Mr. Armes says, was “anyone can do 3-D, but can they do it in time and in volume?” The answer seems to be yes: Today, the firm is wrapping up work on a major Warner Brothers action film, one of its first big start-to-finish conversion jobs – and proof positive that it’s popped out from the crowd.