A few months ago, I had posted a video on this very website about software developed that could analyze a single photograph or even a painting, and derive a 3D model from it. It has been the single greatest attractor of visitors to this website, totaling over 50,000 of the 108,000 or so hits this site has seen since I started tracking stats in February.
It seems that developer of this insanely cool software has kept busy. He has more recently developed software that (I can only assume) uses some of the same techniques as the original software but now, not only does it create the 3D model, it can also create incredibly realistic mouth movements. A lot of you may be thinking “What the hell? Why do I care?” Well, I want you to watch any computer animated movie or TV show and tell me what you think is the most un-realistic part of the animation. Sure, the hands could usually use some work, but the mouth movements are nearly almost terrible.
I’m no computer artist, so I can’t tell you how complex the process of animating speech or even just the act of smiling is, but judging by the work that I’ve seen over the past 10 years, it must not be easy. I mean, in a human face, there are hundreds of bones, muscles, tendons, and other stuff that all work in tandem to produce smiles or frowns, so when an animator has to recreate these emotions, they have to approximate. There’s no way they’re going to actually recreate every muscle, tendon, piece of flesh, and create real-life physics; it would just take way too long.
That’s where this software comes in. It’ll look at a single picture (it doesn’t even need to be a 3D model; it’ll generate that on its own) and allow you to select the movement with sliders for simple emotions like smiling or frowning, or with more complex scripting for speech. Even if the speech takes a bit longer to create than a single smile, it’s still amazingly impressive.
Leave a comment with your thoughts.
One Comment
Permalink
Beautiful, both literally and on a conceptual level. For an idea of just how much of a feat this is, check out some of the most sophisticated stuff going on in computer vision right now
Using Data to “Brute Force” Hard Problems in Vision and Graphics - http://video.google.com/videoplay?docid=-8639996003880499413
Microsofts Photosynth -
http://video.google.com/videoplay?docid=-5485126614666667267&q=photosynth&total=103&start=0&num=20&so=0&type=search&plindex=0
Both examples require the use of vast data sets to ‘tear’ the object from it’s two dimentional plane into a third, never mind the horrendous task of creating a simulation of muscles tendons and underlying bone structure from the derived model.