Could only just tell but it drew me in, very convincing, wonder what kind of tesselation would be required to have a full range of facial movement like that because I couldn't even make out the wireframe, looked very dense.
It's the best I've seen, but something about the mouth interacting with the teeth is what seems to be lacking I think. Something about it still feels creepy to me though. I mean if I were actually talking to this girl face to face it would freak me the fook out.
Finding faults doesn't make you cooler, guys (I wonder what 'faults' people would pick out in the original video). That was seriously awesome, but I'd like to see more practical applications and implications of it (re-rendering videotape isn't very useful).
right off the bat, it still sinks in the valley for me. The eyes and mouth do not move correctly at all. Totally unnatural. The face material also looks like rubber, or like it's coated in a pound of makeup. Ugh.
Still a ways to go...
My wife just watched it and said, "Computers are going to be used for bad things..."
Then she asked me if I didn't already know it was fake, would I know it was fake- I don't think I would. Only reason I was fixated on the mouth was because I was looking for telltale signs - if I was watching casually I don't think I'd notice.
They did a good job of recreating all those endearing little lip curls/pursing.
Nobody said it does but you can't say 'end to uncanny valley' and expect everyone to bow down and kiss your feet when there's still a long way to go. Is it cool? Yeah it looks very realistic, mostly because they camera matched the face to the real girls hair and clothes, which help sell the illusion, but the downfall is still the eyes and the mouth wonkiness. It doesn't look generations better then the stuff in Beowulf. I think for the average user/person they may think it's a real person, those of us in the industry and seeing and trying to do the same thing we pick out the flaws fairly quickly.
There were some perspective/distortion issues, and before even playing the video, just the video thumbnail, I could spot that it was another 3d character (or, partly that is).
so they basically traced this over a real person's face, and only the face. not even close. but nice try. they almost had it due to Youtube's crappy compression. but the mouth and eyes give it away.
Yah, would have been nice to see the technology demonstrated in a more creative way. At the same time, I can see why they took the "I can't believe it's not butter!" approach to promoting it.
well that was alright, the mouth I think particularly the lower lip and the way the teeth dont seem to be part of the lip animation looked a little strange but i think alot of people could be fooled if they werent told. That said it isnt a very creative example of 3D at all.
Oh, it's just the face... that's kind of a letdown.
But anyway, I don't know for sure how people work with 3D scanners, so I have a question...
(don't think it's worthy of a topic all for itself so I'm asking here)
When they scan a person's head/face, they use it as a base for a simpler mesh, of course.
But when they have to animate the face... they do it all by hand with no 3D reference at all?
I mean, if they don't... wouldn't be easier to do more scans with the actor making different expressions? Open mouth, the "Ohs", the "Oos", the "Ees"... Smiling, frowning and all that.
And they say if they were to do it again they'd need 6 months? I'm with you, I'm not blown away by this. They only did the face, and only show it from the exact same angle as the video they recorded, under not very challenging lighting conditions. That the face looks good is thanks to Paul Debevec's scanning system, which we already know is excellent.
They said that creating the face and setting up the rig would take them 6 months, the actual animation side of things has a short turn around. Their primary product is the animation and face tracking system, not actually creating and rigging the mesh. Though 6 months for a single face is a hefty amount of time.
Looks pretty cool though. I think it would have been more effective if they had just left their CG face to do all the talking without the comping rather than doing the short render pass bit at the end.
Replies
impressive!
Urrh.. I was gonna post that I think the eyes are weird in a few shots but hey.. it's probably only what I think cause I know it isn't real.
Isn't that what they used for the UT3 facial animations too?
Still creepy.
http://www.awntv.com/videos/image-metrics-emily-project
Making of clip:
http://media.fxguide.com/fxguidetv/fxguidetv-ep030.mov
And it's only the face the hair/body are actually the girls and only the face is cgi.
That's exactly what I thought when I saw that. All she needs now is a little black drool to get the whole package going.
Still a ways to go...
Then she asked me if I didn't already know it was fake, would I know it was fake- I don't think I would. Only reason I was fixated on the mouth was because I was looking for telltale signs - if I was watching casually I don't think I'd notice.
They did a good job of recreating all those endearing little lip curls/pursing.
Nobody said it does but you can't say 'end to uncanny valley' and expect everyone to bow down and kiss your feet when there's still a long way to go. Is it cool? Yeah it looks very realistic, mostly because they camera matched the face to the real girls hair and clothes, which help sell the illusion, but the downfall is still the eyes and the mouth wonkiness. It doesn't look generations better then the stuff in Beowulf. I think for the average user/person they may think it's a real person, those of us in the industry and seeing and trying to do the same thing we pick out the flaws fairly quickly.
exactly, everything was real but the face. :poly121:
Impressive, but not perferct.
[ame]http://www.youtube.com/watch?v=dMwhZryRUr4[/ame]
But anyway, I don't know for sure how people work with 3D scanners, so I have a question...
(don't think it's worthy of a topic all for itself so I'm asking here)
When they scan a person's head/face, they use it as a base for a simpler mesh, of course.
But when they have to animate the face... they do it all by hand with no 3D reference at all?
I mean, if they don't... wouldn't be easier to do more scans with the actor making different expressions? Open mouth, the "Ohs", the "Oos", the "Ees"... Smiling, frowning and all that.
Looks pretty cool though. I think it would have been more effective if they had just left their CG face to do all the talking without the comping rather than doing the short render pass bit at the end.
-caseyjones