Uncanny Valley

Apologies if this has already been posted, but I just came across this and was amazed. This is an animation:

http://ca.youtube.com/watch?v=bLiX5d3rC6o

Good stuff. But that's not CG animation. That's live action with a CG face driven by markerless mocap facial data. The part from 1:18 to 1:31 shows the superimposed facial data on the live plate. The point is to show that they don't need markers to capture the live data and drive the facial rig.

http://www.youtube.com/watch?v=SwAV2fXoy6E&eurl

Someday...

This type of stuff is going to affect Hollywood more than it will gaming. The studios will finally have an actor that can never grow old or cause trouble with their personal lives. The human who controls the actor can easily be replaced with a new person. Voices can be altered. Maybe this will get rid of the need for a good looking face and make acting based purely on acting ability. The big problem was the 'uncanny valley' and now that's gone.

As far as SL goes, I think I'd get tired of making faces at my computer.

No, I think they still have the Uncanny Valley problem. The eyes don't focus in the sample they gave. I mean, it's pretty close, but it's far enough off to give you that uneasy feeling that something's wrong. Not a leap forward, just an incremental improvement to an existing process.

For a better example, go to their website: http://www.image-metrics.com/

I presume we will see this in games in just a few years judging by their client list. Impressive.