If you're not already familiar with it, here's a term used you'll see a lot of when people talk about computer graphics (or even robots): The Uncanny Valley. Learn the definition and you'll be able to impress your friends at parties. It works like this: if you were to draw a graph of people's emotional responses to things, it would follow a clear pattern. If something doesn't resemble a human at all (say, a picture of a rock), people have no response to it. But as you add more and more human-like characteristics -- eyes, noses, smiles, etc. -- people will develop a warmer emotional response to it. You'll start to feel empathy. The response is even greater if the character is animated and moves in realistic ways. But, the graph isn't a completely smooth curve! At some point, the character looks real but not real enough, and suddenly people will actually have the opposite reaction to it. A negative response. The character will look "creepy" or "weird" or "strange." Instead of seeing the human-like characteristics, you only notice what isn't quite right.

Animators have known about this phenomena for decades, and in the computer games industry it really became a concern more recently as graphics became good enough to make photo-realistic characters. Have you noticed that characters who might look okay in screenshots just seem wrong when you play the game? Creepy, waxy, jerky and robotlike? Yep, those chracters are sliding into the uncanny valley: they're human, but there's something wrong with them, and your instincts react negatively. It's fun to compare the characters of EverQuest II with World of Warcraft: WoW's designers consciously made their characters look like cartoons, exaggerated in dozens of ways, making them less human but high on the curve and easy to empathize with. EverQuest II went for a more realistic look, and plunged into the valley as a result.



I dwell on this topic as a followup to yesterday's blog about id's Tech5 Software. We now have the ability to render, in real-time, startlingly human faces. Every pore, blemish, freckle, and eyebrow-hair -- it's all on the screen!

But will the characters look right? That's the next step, and it's going to require animation that's as solid as the rendering. I suspect that will be the focus in the next five or ten years of videogame graphics development: progressively better and more realistic animations. It's going to require moving beyond 'canned' animations and into 'procedural' animations, blended and rendered on the fly.

A lot of work is going on in this space. My favorite example is the new Indiana Jones game, which uses NaturalMotion technology to dynamically animate the characters. I saw the demo last year at E3 and was really impressed: individual characters would do their best to keep their balance amidst all kinds of forces, like standing on a moving streetcar while being punched in the face. Or staying on a wriggling rope bridge while boulders landed on it. Admittedly the characters looked a little like marionettes being smacked around, but the technology was a real blast to interact with. And it's still early! We're going to see more and more technology focused just on realistic (and dynamic) animations.

At this year's SIGGRAPH (Computer Graphics) convention, EA's chief visual and technical officer gave a keynote speech that raised some of these same points. You can read a summary of the keynote on GamaSutra, where he talks about what it takes to move beyond the valley.

And of course after that comes improvements in AI... but that's a blog for another day.

      -Fargo