Inside Epic’s incredible first attempt at real VR gaming

Walking through the darkened hallways of Epic Games, I know three things about the secret project I’m about to see.

The first is that it bears the promisingly straightforward title Bullet Train. The second is that it’s an experience built for Oculus Touch, arguably the most advanced motion control system that ordinary people stand a chance of using. The third is that it’s the biggest step Epic has taken into the emerging medium of virtual reality.

Epic’s VR war room is a small office space strewn with prototype equipment: an HTC Vive, a small stack of Oculus Rift cases, even an old Razer Hydra controller. It’s also nearly deserted. After a week of late nights, the team has just locked down a near-final build. They’ve got another two days of squashing bugs before they can hand it off to Oculus, and a week before it appears for the first time at the Oculus Connect convention in Hollywood. But it’s good enough to start showing off.

Senior designer Nick Donaldson was in the office until after midnight, and he seems both weary and animated as he copies over the latest build of Bullet Train, helps me into a Crescent Bay headset, and hands me two half-moon controllers. It’s only the second time I’ve used Oculus Touch, and I’m still feeling my way around the buttons when the action starts.

Oculus’ only Touch demo so far is called "Toybox," and it’s exactly what it sounds like. A virtual partner talks you through playing with things like action figures and slingshots, and the most high-stress part is shooting toy missiles from a miniature tank. Bullet Train, as its name suggests, drops me in a fast-moving subway car that has me turning around to get my bearings — I do a quick double-take at a couple of reassuringly familiar MTA etiquette PSAs. A small arsenal of guns appears on the seat beside me, and a disembodied voice tells me that I can teleport while slowing down time. Also, apparently, a lot of people want to kill me.

As shooters go, this is a simple setup. In VR, it feels fresh and foreign — there’s a real sense of threat as the train screeches to a halt. Where games have trained me to think in terms of hitting keys and buttons, I start paying attention to things like which hand to use first when picking up a two-handed rifle. As the train’s doors open, I teleport behind a row of masked goons on the platform, avoiding their hail of gunfire. Without thinking, I get on one knee to duck behind cover. I can only move a couple of steps in any direction, but every time the goons start getting too close, I blink to another of several possible locations. When one of my guns runs out of bullets, I toss it away like a disposable razor. I can even dodge bullets.

But the world of Bullet Train isn't precisely the Matrix. This is one of the most immersive and advanced VR experiences I’ve tried, but I’m still not quite interacting with it. My virtual hands move almost exactly like my real ones, but they interact with the world like a pair of sticky pads. A rifle will automatically leap between my fingers if I grab one of two highlighted sections, and I can’t just shift it from hand to hand — I have to awkwardly drop it and start over. The fact that everything is nearly weightless starts to break illusion that I’m really holding a weapon. Even so, the final boss — a giant robot shooting missiles that I have to pluck from the air and throw back — comes all too soon.

When I take the Rift off, I look over at Donaldson’s computer, where the Unreal editor displays a God’s-eye view of the train station. It looks just like any other FPS game map — just rendered on a different kind of screen. "I think building a 2D game and building a 3D game aren't actually that different, other than the extra dimension," says Donaldson. It’s the little things that separate a good VR game from a disastrous one — especially anything that could induce motion sickness.

This is one of the most immersive and advanced VR experiences I’ve tried

That’s one of both Oculus and Epic’s main preoccupations. Bullet Train uses teleportation because it’s one of the safest ways to move in virtual reality (putting players in a vehicle or using a distant third-person camera are others). Even plans to add a rideable escalator were quickly squelched. "We sent this to Oculus and they're like, 'Mm. No, nope, nope, nope, nope!’ They felt queasy on the way down," recalls Donaldson. The entire game is designed to carefully position players in the right spaces, looking the right direction — the teleporter will only move you to predetermined spots, placed carefully behind trash cans or desks oriented toward the center of the map.

Players don't necessarily expect photorealism from VR; because it's so graphically demanding, an experience like Bullet Train has to use simpler graphics than one of Epic's 2D demos. But things like scale, camera height, and sound placement matter a great deal. "This isn't some abstraction anymore," says Donaldson, when they get it right. "This isn't a screen on the wall. I feel like I'm actually there."

As Oculus prepares for its consumer launch in 2016, the big question is whether VR is actually here. But we’ve been asking that question for so long that it’s starting to feel irrelevant. Yes, it’s here. The bigger question is this: does Epic (or any other company) have any idea what it should look like?

When Epic lead programmer Nick Whiting agreed to check out a weird little prototype from a company called Oculus, he didn't expect it to slowly take over his life. His friend Nate Mitchell had just walked away from video game interface company Scaleform to take a gamble on a teenage inventor named Palmer Luckey — and the project he said would bring virtual reality back from the dead.

Today, Whiting is video-calling from his home base in Seattle, along with studio manager Ray Davis. Though he hasn’t pointed it out yet, they’re sitting in front of a small dent, one of the various pieces of minor property damage caused by accidentally punching the wall during testing.

"They said, 'Hey, we have this crazy idea for a VR headset,'" Whiting says. "'If we send you one, will you hook it up to [Unreal] and see what it looks like?'"

The headset, as Donaldson remembers it, wasn’t much more than a taped-up screen. "There was some Doom level or something that they were rendering in there, some horrible smeary frame rate, and I was like, "Wow, this is awful! This is amazing!" he says. "It was terrible, but it was so good! It was exactly what I dreamed of — being in those 3D worlds."

Oculus was taking an early version of its headset to the 2013 CES, and it needed something to show. Whiting agreed to port over Epic Citadel, an Unreal Engine 3 demo developed for iOS in 2010. CES attendees were as impressed with the tech as he had been. The Rift won a half-dozen awards (including "Best in Show" from The Verge), and Epic Citadel marked the beginning of an ongoing collaboration.

Epic phased in support for VR in Unreal Engine 4, and it created or adapted flagship demos for each new version of the Rift: a miniature lava fortress called Elemental for Crystal Cove, a two-person sword-fighting game called Couch Knights for the DK2, and a street battle called Showdown for Crescent Bay. But the projects were near-solo efforts, produced under strict constraints. Showdown, for example, repurposed material from old non-VR demos, slowed down drastically to disguise the fact that there was only 6 seconds of motion capture material. It was made by Donaldson and a single artist over five weeks.

Bullet Train took twice as long and was created from scratch, with around a half-dozen Epic designers and artists lending their time to the project. Whiting and Donaldson led it from opposite sides of the country — Donaldson works out of North Carolina, Whiting from Seattle. Their collaboration process involves hours of Skype video calls with frequent jumps into virtual reality for testing. Donaldson often doesn't even bother to take off the Rift, flipping it up like a welder's mask to look at his monitor.

VR development occasionally sounds like the kind of thing that sends science fiction protagonists into existential crises. "After a day of heavy VR, I get this weird reality effect," says Davis. "It's like you're now becoming more... scrutinizing [of] reality or something." Ordinary objects have to be constantly examined, tweaked, and tested. "In our experience, you spend a lot of time looking at your hands, right? In reality, you tend to not focus on your hands, because why would you?" he says. "I think we were debugging some of the Touch stuff or something, and I was going to the elevators and was like, ‘Oooh, I have hands! And look how they move!’"

But the real problems are comparatively mundane — including finding the cutting-edge VR they need to create these kinds of demos in the first place. Until the real mass-market product comes out next year, Oculus can't produce enough Crescent Bay headset and Touch controller prototypes to meet demand, so Epic's developers only have access to a couple of each. And that's assuming they haven’t sent one of the delicate hand-built designs back for repairs. "We're constantly breaking things, crushing cables," says Whiting. "I think I actually launched one of my Oculus Touch controllers across the room this morning," adds Davis. "But the Touch controller survived!"

We're constantly breaking things

Whiting also had to convince the rest of the company that VR was a gamble worth taking. Epic founder Tim Sweeney was already making games during the first wave of virtual reality, which ran from the late 1980s to the mid-1990s, and he wasn’t impressed with what he saw. "He was a non-believer for a little while," says Whiting — until, according to Davis, he tried out one of Valve's room-sized VR demos, at that point far more advanced than Oculus’ test units. "He and [Epic engineering VP] Dan Vogel went inside and were like, 'Oh, this is pretty cool. This is much better.'"

When I ask Sweeney about it, he recalls 20th century VR less than fondly. "You'd have two little 320 x 200 displays. The pixels would be like the size of a quarter sitting on a table, and running at 20 frames a second, it was just a completely disjointed experience," he says. "It was always like, ‘I have this thing on my head, and I'm looking at some monitors, and it's all pretty crappy.'"

Today, Sweeney is an enthusiastic convert — VR, he insists, will change the world. "I think it's really hard to remain a skeptic right now," he says. "It's not going to happen in the next year or two, but it's going to happen in the next decade. It's going to have a very pervasive effect on everything that we do." Early VR headsets were the PalmPilot to the Rift's iPhone. The idea was there, but the technology wasn't.

The smartphone analogy comes up a lot in the VR industry. It aptly conveys two pieces of common wisdom: virtual reality is going to transform computing as we know it, and 90 percent of what early adopters see will be awful. "In the early days of iPhone, every game had virtual D-pads and they were trying to simulate a mouse and keyboard. Occasionally it would work, but more often than not, it was terrible. I think that's where we're really looking at VR," says Whiting. "I think the real magic is about two to three years out, when people really wallow in the space and learn it."

But VR has more hype to deliver on than smartphones did, especially because the greatest excitement is reserved for things that don’t exist yet — few people are totally satisfied with the goggles-and-controller technology that’s available right now. Like many other early adopters, Epic’s VR enthusiasts predict that virtual reality will merge with augmented reality like Microsoft’s HoloLens, allowing wearers to alternately block out the world and project visuals onto it. When Sweeney says "VR" will change the world, that’s what he’s talking about.

"I think of VR as having a potential audience of like, 250 million people," says Sweeney. "It's not all of humanity. It's not billions of people, like an iPhone." But when you can get a pair of stylish augmented reality Oakleys, that’s a different story. "At that point, the audience is maybe 4 billion people. It will revolutionize everything and, I think, displace all existing computing platforms, including PCs, and smartphones, and tablets, and everything else." That includes Epic’s own desktop game-making software. "Picture every sort of content creation tool: Photoshop, 3D modeling, 3D editors associated with games — like the Unreal Editor," says Sweeney. "Over time these will evolve to be VR-based. There's no question that people creating content for VR would want to create that content in VR."

For now, though, the vast majority of experiences are for entertainment only. At Epic, entertainment has traditionally meant games, whether that’s a series like Gears of War or one of the many outside projects made with the Unreal Engine. But in recent years, the company has broadened its scope. It hired a Lucasarts executive (and veteran special effects artist) named Kim Libreri to serve as CTO, and it started putting out demos that felt less like first-person shooters: a Pixar-esque animated short about kites, a virtual-reality Hobbit tie-in produced with special effects company Weta Workshop. Bullet Train’s environments were created by its newest Hollywood hire, a digital artist named Jerome Platteaux.

Sweeney has described the Unreal Engine as a "common language" for a world where film and gaming converge, and there’s nowhere that convergence is more strongly felt than in VR. "I think we see almost equal amounts of game companies and movie companies using the engine for producing content at this point for VR," Libreri tells me. Right now, VR film gravitates towards live-action documentaries and short film or TV tie-ins — the series Sleepy Hollow was just awarded an Emmy for its virtual reality experience. While they give viewers an unprecedented amount of control compared to traditional video, these experiences are still essentially passive. But Libreri thinks that they’re about to start feeling a lot more game-like.

"The whole idea of just looking at something becomes a little bit dull," he says — once audiences can use headsets as a matter of course, they’ll want more interaction. "Once you’ve actually been able to move something in VR, you become pretty dissatisfied. No matter how beautiful and awesome flying over the Grand Canyon is, you still want to be able to control your own destiny." And now that headset companies like Oculus have enough faith in their control systems to show them off in public, that’s finally starting to look like an option.

Epic won’t say more about its plans beyond Bullet Train, except that new projects will start "as soon as we're done." It’s unclear whether the company will get a dedicated VR division, instead of pulling together a handful of people for each project. And a complete game seems far away. Epic’s developers can describe lots of ideas for virtual reality, but if they’re ready to embark on any grand, unified plans, they’re not telling. Their creations are designed to showcase specific features on each of Oculus’ new prototypes: 360-degree head tracking for Showdown and Crescent Bay, interactivity for Bullet Train and the Touch controllers. That’s likely to hold for whatever technological leap comes next — Whiting and Davis hope it’s wireless headsets. "We haven't actually started thinking about the next steps, since we're just trying to finish this step," Donaldson says.

For now, people will be able to try Bullet Train at Oculus Connect, starting today. It’s hardly the fast-paced arena-running of Unreal Tournament, but after years of hands-off demos, it’s probably the most classically Epic-style VR experience yet. "It may not be for everyone. You're shooting people, you're teleporting, you're slowing down time. It's a little bit insane," says Donaldson, his exhaustion apparently forgotten. "But it's fun. And that's what we wanted to make, because we're a game company."

Read next: Our Oculus Rift review