Nice technology; bad video to sell it with. If they wanted the viewer to be shocked at the end ("You mean, she's not real?!?!"), they failed. They didn't make the point clear, by just slapping what could be assumed to be overlays on her face. They could've at least deconstructed or melted her or something.

...or reveal at the end that it's a 200lb programmer with stubble whose appearance was modified into "Emily" and voiced dubbed.

Oh, come on, that was far, far better than an aweful lot I've seen, it had me for a little bit.

Of course, that's helped by the facts that:
* This is low res YouTube we're talking about, hides a lot of problems
* They only overlaid a face on top of real video, so all the issues of body movement and realism of other objects was mute. (It helped a lot, as I was looking at the body going, hmm, no, that looks pretty real)

Still, that's some of the damned best mouth movement I've seen on a cg character, mocapped or not.

Although, then when you go back to their full cg characters, you fall back to the same issues there's always been, as on the front page here.

It shows that having as much as possible being real helps an awful lot in selling cg

Oh, man, and I hate that 'real journalists' can spout crap about stuff like this.

this article states that everything you see in the video is CG, which is patently false

Very misleading / ambiguous video. What parts of that scene were CG? None except for the few overlays, right? Big deal, at that resolution and without realistic rendering we can't tell how good it is.

We've seen Emily before, and they're getting better at it! http://www.videosift.com/video/The-Uncanny-Valley-is-nearly-here

>> ^rychan:
Very misleading / ambiguous video. What parts of that scene were CG? None except for the few overlays, right? Big deal, at that resolution and without realistic rendering we can't tell how good it is.


I'm pretty sure the whole thing up until they show "the real Emily" is CG.

>> ^rychan:
Very misleading / ambiguous video. What parts of that scene were CG? None except for the few overlays, right? Big deal, at that resolution and without realistic rendering we can't tell how good it is.


Follow this link for a higher rez video:

http://www.awntv.com/videos/image-metrics-emily-project

at this rez you can see the seams of the animation, but still very believable.

>> ^therealblankman:
We've seen Emily before, and they're getting better at it! http://www.videosift.com/video/The-Uncanny-Valley-is-nearly-here


Yeah, that video is from the front page of the company's website, check it out there for a higher rez
version:

http://www.image-metrics.com/

The "real Emily" looks exactly the same as the CG Emily so far as I can tell. Is that what's impressive about it?

>> ^SaNdMaN:
>> ^rychan:
Very misleading / ambiguous video. What parts of that scene were CG? None except for the few overlays, right? Big deal, at that resolution and without realistic rendering we can't tell how good it is.

I'm pretty sure the whole thing up until they show "the real Emily" is CG.


And what possible reason do you have to say that? I mean, WHY do people insist on saying something that they cannot back up with ANY reason?

The part where they show the face in cg overlay... THAT'S it for what they've done to the video, overlaid a CG face onto her face after using the very video they're sampling from to get the movement. Why would they show that being the only cg part if they did everything? They'd be shouting from the rooftops if it were everything, they'd deconstruct everything in the room if it were really ALL CG.

Plus, the rest of her body etc. LOOKS more real than the face... especially the mouth.

It's still darn impressive.

But again, it shows how much more you can get away with if you surround your cg in reality.

It's quite impressive, but having all the rest being real helps it enormously

^ What the heck are you talking about? The whole point is to surprise people that the person they've just seen explaining the technology is actually CGI. Did you think that the CGI demonstration was just the face turning red and black? That would be stupid and pointless.

these people are wasting their time with duping emily.. they should be putting famous actress faces on there

Comment hidden because you are ignoring dag.(show it anyway)

It could lead to actors in production for multiple movies at the same time, just dub in the voice later.

Actors will be all about licensing their image instead of actually having to show up.

Once Angelina Jolie's scan hits the black market - a new porn era will begin.

I could tell almost immediately that it was not real. Her upper lip was too stiff, and the skin and muscles near her nose and below her cheek bones were too rigid. There were no natural creases around her cheeks when her lips curled. You can see the effect when the Real Emily curls her lips - heavy crease lines flanking her upper lip and on the edges of her mouth.

FAAAAAKE!

Slap a shiny torus on her face and toss her up on DeviantArt!

>> ^SaNdMaN:
^ What the heck are you talking about? The whole point is to surprise people that the person they've just seen explaining the technology is actually CGI. Did you think that the CGI demonstration was just the face turning red and black? That would be stupid and pointless.


The demonstration was showing that THE FACE was CG for the whole video.

Geebus.

They are showing that their motion tracking system is really great at FACES.

The rest is REAL.

You want proof? how about this?:

Using (USC's) Institute for Creative Technologies' special scanning system that can capture facial details down to the individual pore, the face of actress Emily O'Brien was transformed into a digital representation of herself, which could then be entirely machine-manipulated. A special spherical lighting rig captured O'Brien in 35 reference facial poses using a pair of high resolution digital cameras. The facial maps were then converted into 3D data using Image Metrics' proprietary markerless motion capture technology.

Lame; still deep in the valley for me... they should have fit other small women's faces on her head.

That would have been impressive.

ugh, definitely in the valley and definitely closer to coming out, which means really, really offputting. right from the start, there was something very, very slightly and yet horribly off. uncanny valley is such an interesting phenomenon...

her eyelids blink too slowly and too uniformly, as in the same speed up and down.

To me the amazing thing isn't really how realistic it looks, but how there is still definitely something uncanny about it. Its truly weird and fascinating to be able to notice every single minuscule discrepancy in the cg so effortlessly.

It reminds of a video about neurological phenomena, specifically a disorder in which the ability to recognise faces is impaired. Sufferers are incapable of recognising their own mother, but more than that they are often repulsed and distrustful of them because the brain thinks the face it sees is an imposter. (I think sufferers can still recognise faces, but the link pairing facial recognition to memory is damaged, so they see a familiar face, but don't recognise it as their mother.)

Anyway, I suppose our ability to recognise an imposter isn't as bizarre as our repulsion to the imposter. Why have we evolved to be quite so wary of things trying to be human?

"Uncanny Valley" is a consensus theory but not scientifically proven.

Show us your boobs!

fake! it's real!

>> ^quantumushroom:
"Uncanny Valley" is a consensus theory but not scientifically proven.


Except for the scientific papers that measure and confirm it, you mean?

For instance, http://www.bartneck.de/publications/2007/uncannyCliff/index.html

There's other publications I could point to:
# Hanson, D. (2006). Exploring the Aesthetic Range for Humanoid Robots. Proceedings of the CogSci Workshop Towards social Mechanisms of android science, Stresa.

# MacDorman, K. F. (2006). Subjective ratings of robot video clips for human likeness, familiarity, and eeriness: An exploration of the uncanny valley. Proceedings of the ICCS/CogSci-2006 Long Symposium: Toward Social Mechanisms of Android Science, Vancouver.

The original paper is "Mori, M. (1970). The Uncanny Valley. Energy, 7, 33-35." by the way, but it doesn't include empirical studies.

51 votes!?
Seriously, come on! How can you guys support this sad software-co marketing peddling video?

How can this software be so much better than that which exists already?

• Motion capture in everyday surroundings; and
  
• Motion Capture without markers, Awesome!