We're getting tantalisingly close to crossing the uncanny valley - the chasm between what we can create with our very best CGI efforts and what we actually see in reality
What's fascinating is that it is getting easier to illustrate the problem. As we progress through generations of techniques we can quite accurately plot where any given example sits on the now quite famous curve.
For anyone that's not seen it, it's easy to explain.
Cuddly, toy-like robots don't creep us out. They're obviously not real and we feel OK about them. The more like a real human our attempts get, the more they look like - well, they look like something that lacks whatever it is to be human, even if the look pretty realistic.
It gets even worse with movement. In a way, it's easy enough to make a photo-real model (well, not that easy perhaps) but once it starts moving it will, likely as not, look more like a mechanical digger than, say, a person walking down the street.
The trouble is that the closer you get to imitating reality, the greater the scrutiny you invite. If something looks unbelievably good, it is likely to be unbelievable.
Here's an example that really has us wondering. If it is what it appears to be, then it is a truly remarkable achievement. There's nothing uncanny at all about this. And there's a real twist at the end of the short video.
But while you're watching this video from Image Metrics, remember that you're not seeing an animation. This is motion capture in one form or another. In other words, it's replicating what the real person its modelled on is actually doing.
You could argue that it disqualifies it from the competition because this is not really an animation: this is not a procedural or algorithmic creation; it's just a copy, in the same sense that a photo is a copy of the scene that was photographed.
But nevertheless, what we are seeing here is the creation of a 3D model that you can manipulate quite independently of the source model.
This is, essentially, just a recording. The question is: what it look like if it were to say things never said by the original and to move independently of the source material.
We say: very good, but the jury's still out.
Video after the break