Very impressive, isn't it? Well, I'm not buying it. First of all, the clip says "What you are about to see is a wholly computer-generated animation." This is not true. It's clear that the facial area is synthetic, but everything else is simply a video. In fact, for each of her answers to the question asked, a new piece of film was shot with the actress saying exactly the words you hear. This is a far cry from having the computer come up with realistic motion from scratch.
Second, this really isn't animation at all, it's motion capture. This means that they put a large number of location sensors at points on the actress's face while she spoke, and a computer recorded all of the positions of those points at each point in time. This data was then used to create a computer model of the speaking actress speaking those exact words. Again, this is not a purely synthetic model of her facial movements, it's literally measuring her face while she's speaking. It is not clear that the computer model could be used to say something other than what was said when the actress's face was measured.
Finally, the only really synthetic thing here is the facial shading and color. Actually, this might have been based in part on color and shading samples from the film, but I'll give the creators the benefit of the doubt there. But this benefit is easy for me to give; upon close inspection I don't really think the color and shading look realistic.
Having worked in the computer graphics and vision research world has made me sensitive to what I think are inflated claims of success in these fields. I often see such claims in the popular press, and popular-science press. I recommend taking such things with a grain of salt. I often suspect that these claims are made as bait while fishing for funding, either venture capital or government grants.
0 comments:
Post a Comment