We’ve seen before how modern moviemakers attach sensors to stars’ faces to generate digital versions of the actors. But this is something new, and it doesn’t require any involvement from the victim—er, person—being modeled. As long as there are ample pictures or videos of a person that capture a range of expressions and movement, U-Dub’s scientists can create a controllable computer model of that person’s face all by themselves.
They do it by combining a few different methods: 3-D facial reconstruction, tracking, alignment, and multi-texture modeling. Once the digital face has been constructed, they can paste it on someone else’s head.
One of the more benign applications would be to simulate a younger version of a celebrity, as happens with Tom Hanks about 45 seconds into the video: Tom Hanks at three different ages, saying the same thing. Just a bit eerie.
Right after the Hankses, there’s Ian McKellan on the right speaking the words of Daniel Craig on the left. It’s bad enough with celebrities, but when you get into politicians swapping identities, well, it’s downright chilling.