How Milo met CAVE

I just read an interesting article, a behind-the-scenes look at the infamous “Milo” demo Peter Molyneux did at 2009’s E3 to introduce Project Natal, i.e., Kinect.

This article is related to VR in two ways. First, the usual progression of overhyping the capabilities of some new technology and then falling flat on one’s face because not even one’s own developers know what the new technology’s capabilities actually are is something that should be very familiar to anyone working in the VR field.

But here’s the quote that really got my interest (emphasis is mine):

Others recall worrying about the presentation not being live, and thinking people might assume it was fake. Milo worked well, they say, but filming someone playing produced an optical illusion where it looked like Milo was staring at the audience rather than the player. So for the presentation, the team hired an actress to record a version of the sequence that would look normal on camera, then had her pretend to play along with the recording. … “We brought [Claire] in fairly late, probably in the last two or three weeks before E3, because we couldn’t get it to [look right]” says a Milo team member. “And we said, ‘We can’t do this. We’re gonna have to make a video.’ So she acted to a video. “Was that obvious to you?” Following Molyneux’s presentation, fans picked apart the video, noting that it looked fake in certain places.

Gee, sounds familiar? This is, of course, the exact problem posed by filming a holographic display, and a person inside interacting with it. In a holographic display, the images on the screens are generated for the precise point of view of the person using it, not the camera. This means it looks wrong when filmed straight up. If, on the other hand, it’s filmed so it looks right on camera, then the person inside will have a very hard time using it properly. Catch 22.

With the “Milo” demo, the problem was similar. Because the game was set up to interact with whoever was watching it, it ended up interacting with the camera, so to speak, instead of with the player. Now, if the Milo software had been set up with the level of flexibility of proper VR software, it would have been an easy fix to adapt the character’s gaze direction etc. to a filming setting, but since game software in the past never had to deal with this kind of non-rigid environment, it typically ends up fully vertically integrated, and making this tiny change would probably have taken months of work (that’s kind of what I meant when I said “not even one’s own developers know what the new technology’s capabilities actually are” above). Am I saying that Milo failed because of the demo video? No. But I don’t think it helped, either.

The take-home message here is that mainstream games are slowly converging towards approaches that have been embodied in proper VR software for a long time now, without really noticing it, and are repeating old mistakes. The Oculus Rift will really bring that out front and center. And I am really hoping it won’t fall flat on its face simply because software developers didn’t do their homework.

Please leave a reply!