Sony subsidiary MotionPortrait have been demonstrating their latest 3D technology: automatically creating a 3D graphic from a 2D photo of a person’s face. The system – also called MotionPortrait – can identify the mouth, eyes and other features of the human face, and in two seconds construct a 3D representation of that face which can be manipulated, rotated, or even overlaid on top of video clips.
Video demos after the cut
MotionPortrait expect the system to be used in cellphones, among other things, with the processing done on a remote server. Once the 3D render has been created, it can be edited – adding facial hair or changing hairstyles, altering skin tone, etc – or blended with another avatar. It can even be panned by 40-degrees in each direction, without looking too artificial, which makes overlaying it onto video clips (as you can see below) possible.
While higher quality and more realistic 3D renders are technically possible, the MotionPortrait team have purposefully decided to keep things straightforward for their consumer target audience. As well as cellphones, they envisage the system to show up in digital photo frames, digital cameras, TVs and other embedded devices; it’s set to begin shipping sometime this year.
CGI face automatically overlaid onto real actor:
CGI face automatically overlaid onto pre-existing movie: