Virtual reality is truly an incredible technology that deserves to be in everyone’s hands, or eyes rather. But currently, the devices that allow people to experience VR stand at opposite ends of a spectrum, neither of which make VR really accessible. On the one hand, you have dedicated machines like the HTC Vive and Oculus Rift. On the other extreme, you have empty shells that require a smartphone, like the Google Cardboard and Samsung Gear VR. Interestingly enough, the middle ground between these two might have just been discovered thanks to the efforts of Microsoft Research. Called FlashBack, it is a combination of technologies that could let even less powerful smartphones and laptops handle the load required for a smooth, jitter-free, nausea-free VR experience.
Despite their differences, both types of VR devices require rather powerful computing power than most devices can offer. The Vive and Rift need rather beefy desktops or at least laptops, while Cardboard and Gear VR require somewhat top of the models. That pretty much leaves mid to low range devices out in the cold. Until now.
Microsoft Research’s FlashBack sort of stitches up together existing technologies and algorithms, like compression, for example, to allow even less powerful devices to display VR content without stuttering or exploding in your face. It pretty much functions like YouTube 360 streaming or how some game streaming methods work, by pre-rendering content in order to save precious processing power and time.
Of course that works nice and well for pre-recorded content, but VR is dynamic, at least in the way that what is displayed depends on your orientation or location. This is where FlashBack gets complicated and smart. It is able to serve up the correct frames of the image depending on that location context. In order to do that, it has to basically pre-render all possible frames into a “mega-frames”.
Normally, that would take up an inordinate amount of space, which would defeat the purpose of making FlashBack work on less equipped devices. But there’s another part to the magic: compression. This allowed the researchers to exponentially reduced the file size of a frame from MB to KB and then stored in the device’s local memory, whether the SSD or RAM. Only when the frames are about to be used are they uncompressed to the graphics chip’s own memory. Then end result? A VR experience on a lower end device that feels and looks almost as good as the real thing on more expensive machines.
The researchers are hardly done optimizing this technique before it becomes a truly viable alternative. It will be interesting to see where, or if, Microsoft takes this technology, as its focus is geared more towards Augmented Reality rather than Virtual. And that’s one case where pre-rendering might not work at all.