Microsoft's chief research and strategy officer Craig Mundie wants to show you the hauntingbridge his team has built over the uncanny valley. Employing Kinect hardware and custom PC software, the research team at Microsoft has created an unnervingly realistic new avatar that can handle text-to-speech when combined with a script and can recognize the words in any order.
"This is a way to create a synthetic model of people that will be acceptable to them when they would look at them on a television or in an Avatar Kinect kind of scenario," Mundie toldUSA Today in a video interview. "There's no reason that we couldn't do that in real time by feeding the information that we get from a Kinect sensor, including its audio input and its 3D modeling, spacial representation, and couple that to the body and the gesture recognition in order to create a full body avatar, that has photo realistic features and full facial animation," he added.
This impressive (if not somewhat terrifying) demo is still very much in the prototype phase, however, and Mundie said it would be "some time before we see it show up in products." We're just hoping those first "products" aren't T-1000s.