Impression Pi is everything a VR headset should be — it just can't pull it off
Making a wireless, hand-tracking, position-finding virtual reality headset is tougher than it looks
12/cdn.vox-cdn.com/uploads/chorus_image/image/45966760/DSC03000.0.0.jpg)
“Everybody agrees that at some point, you want the lightest, smallest pair of glasses or goggles, and you want to see your hands naturally, and you want to see your body naturally, and you want to feel like you're 100 percent totally there.”
Oculus CEO Brendan Iribe said this a few months after his company introduced the Gear VR, a headset based around a Samsung smartphone instead of a separate computer. The wireless Gear VR frees users from being tethered to a desk, but it’s hard not to wish it went further — that a “mobile” headset would involve a lot more actual mobility. When Oculus acquired a hand-tracking company called Nimble VR a few months later, the future seemed clear: light, mobile headsets could let you navigate an imaginary world with your physical body, without the encumberance of a controller.
While we still don’t know what might end up in a finished Gear VR, a San Jose startup called uSens is trying to find this future. The Impression Pi is a Gear VR-esque headset equipped with its own set of cameras, which track both your hands and your position in a room. Its hardware is smaller, more convenient, and even sort of adorable. In theory, Impression Pi is exactly what we should be demanding from contemporary VR. But it’s also an example of why we can’t demand that — at least not yet.
Right now, the Impression Pi looks a lot like many new VR headsets: a handsome design mockup and a working prototype made of foam, 3D-printed plastic, or both. In this case, the mockup is a rounded black-and-white box with a slot for a phone and two holes for cameras, giving it the look of a slightly concerned robot. When it’s not in use, a plate fits onto the back, wrapping the whole thing into a self-contained package. It’s still bulky by general gadget standards, but for a headset, it’s relatively compact. The working model, by contrast, resembles the very first Oculus Rift prototypes. Light but blocky, it uses a built-in 7-inch screen and connects to a PC. Instead of the single pair of cameras that will be in the final model, there are three swappable sets for different features, one of which can be stuck to the headset at any given time.
The first set handles simple video, an option also available in the Gear VR — the dual-camera setup is supposed to improve depth perception, although I didn’t notice much of a difference. The second is an infrared hand-tracking system similar to Nimble VR or the Leap Motion. It looks at a small window of space and, currently, detects the fingers of one hand. The software is still very rough — it can track fully extended fingers across the entire field of view, but there’s no interaction with the environment. With some tweaking, though, it could eventually tap into the existing pool of developers working on gesture-controlled projects.
The third set is doing something more unusual. Wired headsets can use lasers or an external camera to track where your head moves. But without special hardware like the Sixense STEM, mobile ones like the Gear VR can only tell how it’s tilted, something Oculus CTO John Carmack has admitted is a problem. Impression Pi flips the traditional tracking setup around: the headset’s internal infrared cameras can "read" the environment and shift as a user moves. For better results, there’s a separate laser projector that shines a distinctive pattern on the wall, providing a firm frame of reference.
Impression Pi promises a lot with this. A Kickstarter video shows a woman walking down a real-life flight of stairs that she sees as a virtual cave floor. (No matter how good positional head tracking gets, please do not ever do this, if you value your life.) In another example, designers scan an environment and add augmented reality elements. In reality, it’s a long ways away. After some calibration, the Impression Pi could shift around slightly as I moved, but it was either barely noticeable or subject to random jumps. The laser patterns worked better, but not by much.
A lot of this seems like a software issue, and the current tech demos are too simple to get a good idea of how it might improve. But even at best, tracking is finicky; John Carmack has said he considered the idea of using an internal camera for virtual reality, but decided it would be too jittery and imprecise. It’s also a potential battery-killer. For the prototype, that doesn’t matter, but on the phone-based model that’s selling for $279 on Kickstarter, it’s a big deal. It’s an even bigger deal for the $359 "master" package, which includes its own integrated, high-resolution screen and an Android-based interface (which doesn’t exist yet). The phone-based headset will support the iPhone and a range of Android phones, so battery life will vary there. Company CTO Yue Fei says he’s aiming for up to four hours of gameplay on the all-in-one headset — but it might be more like two.
Fei says he’s got some major manufacturers already on board, and a new sensor will allow him to combine all the Impression Pi’s functions into one pair of cameras. The phone-based headsets are supposed to ship to backers in November, giving uSens eight months to fix the ergonomics, get six cameras down to two, fit those cameras inside the headset, and release a software development kit. The all-in-one headset is listed as shipping in December, but Fei said there’s no hard timeframe, and it could slip into 2016.
The Impression Pi ties simple pieces of VR hardware together in a way few other companies have done. It’s trying to fix the Gear VR’s biggest flaws: its awkwardly not-quite-portable hardware, its limited control options, and its total lack of positional tracking. The problem is, those flaws exist because they’re really hard to overcome. Virtual reality of the ‘80s and ‘90s was full of big ideas that were way too advanced for their actual hardware. The Impression Pi might be one of our first modern examples. It’s an interesting project, but if it’s going to be a fun or useful one, it’s still got a long ways to go.