If you’re (still) unfamiliar with the uncanny valley theory, you can get familiar here.
With that out of the way, now you can take a look at Trusim and see the stuff they got going over there. Very, very neat. As it was pointed out to me while discussing it elsewhere, this is not very much new tech, but rather a very good amalgamation of a few existing techs that are out there now, along with very good tools to get the juice out of those techs.
Still, it’s a little glimpse at the kind of thing we’re gonna be seeing in the years to come in many of our games. Which, yet again every time one of these little leaps comes along, will reopen the whole uncanny valley debate.
Brief thoughts follow.
There are three basic ‘schools of reaction’ around the issue or realistic human representation in games. The first one is all for it, all for further refinement and iteration, until a point is reached where the representation is essentially indistinguishable visually from a real human. There is a certain logic in this train of thought, and it makes sense to aim for that, because if we don’t we’d essentially be a little false with ourselves; we can’t be fine with the idea of increasingly better graphics as long as we’re not trying to represent a human. Better graphics include character representation by default.
Another school pays a lot of attention to the uncanny valley theory, and essentially states that “perfection” (which believe it or not, in this issue turns into a very subjective term) is not really necessary. That we should aim to achieve a state of representation which simply makes sense and is realistic enough. Basically, stopping just before we start sliding way down the slopes of the uncanny valley. This thought doesn’t aim for visual perfection, but rather a conceptual perception – as long as the characters make perfect sense amongst themselves and their surroundings, and they are believable enough, then they are conceptually ‘complete’, functional, and do their job just fine. An indistinguishable replication, while nice in theory, runs the natural risk of falling into the uncanny valley and not coming out.
Along with these two, a minority (sizeable as it is, though) either does not care about this because their enjoyment of a game passes through other areas, or borrows concepts and positions from the other two.
Myself, I don’t have a final answer for the poser of the uncanny valley. While initially I’d be all for just ramming through it, speed through the slope and finally reach the holy grail of representation at the other end, the idea of stopping right before the slope does have a certain appeal to me as well. So I usually ask myself the big question: Do we really want to reach the point where our models and avatars are visually indistinguishable from the real thing?
I can’t ask myself what the payoff is, because I know it. It’s the same payoff we’ve been getting for years every time our graphics got better and more complex – that is, we improve graphics for its own sake, and because we like that evolutionary process. The question changes then. I ask myself: Other than perfection for its own sake, an idea that I love, what else is there at that final point?
Not so deep down I must be a terribly romantic guy, because I can’t help but to think that it’s okay to have imperfections, and its okay not to have truly realistic characters as long as we like and we’re happy with what we have. But hey, let’s see what you think.