14 comments

[ 1.6 ms ] story [ 40.9 ms ] thread
And we've been working successfully in this area for quite a while, with a solution you can access today: www.3D-Avatar-Store.com.

We perform photo-realistic 3D Reconstruction given as little as one photo of a person. Given more photos and our reconstruction quality improves. We’re neural net driven, so we’re super fast: original reconstructions in under a second, and refinements require 40% less time. Multiple geometry outputs, including Maya performance-animation rigs, full body skin-tone corrected texture maps, perspective distortion correction and more.

We're exhibiting next week at the Game Developer's Conference, booth #935. Come visit and have an avatar of you made!

Automatically generating the avatar is very cool, but in terms of realism, there is no comparison between what I've found on your company's YouTube channel and the video from the article.

My conclusion is mostly based on this video from your YouTube Channel: http://www.youtube.com/watch?v=iggXtQ-jBBU

...and this video from the article: http://www.youtube.com/watch?v=5d1ZOYU4gpo#t=527s

Finally, this video from your channel, even though it's a few years old, is both technically impressive and downright horrifying. http://www.youtube.com/watch?v=UFS05tR9028

If you have better material to show, I'd recommend featuring it somewhere on your main website.

Really impressive stuff, especially considering that is still just the beginning. The skin around the mouth and cheeks seems completely (uncannily?) real, although the lip movements could still use some work.

Also, whether it was intentional or not, as a Caprica fan, I greatly enjoy that the talking head is named "Zoe". :)

Still deep in the uncanny valley in my opinion.
Exactly my reaction. Beautiful actress but it still gave me the creeps.

I think the approach of decoupling the eyes/mouth/etc. is a mistake. Our brains are trained to unite all of them and when they don't match up naturally it sets off alarm bells in our minds that something is not right.

I think you didn't watch the second video - which was better than the first.
Thank you -- that's what happened to me. I got seriously creeped out by the first video and wrote off the headline as misleading.
You're right. Watching the Nvidia one now.

EDIT:

The Nvidia one was definitely less creepy but I wonder if it can really be used to say anything. It was a more limited demo.

But it did seem to support my hypothesis that decoupling the eyes, etc. was a mistake because that one is using whole facial expressions together. Or at least that's what I gathered.

Agreed-- take a look at the second half of the second video, though. It's a lot better.
Notice how he has no hair. Still a long way to go, I think.
Notice that the model, before texturized, has fang's for teeth... scared.
As we get closer to being able to perfectly replicate the sounds of human speech, I wonder if we'll find an "uncanny valley" effect. Check out 0:10-0:13 in the video (especially the word "now"). I actually prefer this GLaDOS-esque inflection to something perhaps closer to human-producible phonetics. The imperfection in sliding up and down tones is actually very musical, almost like "grace notes".
I wonder how far off all digital characters we are. Seems logical as a next step and to be honest I'm not sure I personally would feel any different about watching a movie with zero flesh and bones humans in it.