Mass Effect: Andromeda has been catching a lot of flak for its facial animations in the past few days, and by all rights it should be. They are off in a fundamental way that just makes them creepy. For the most part they are ok, but for a few characters (like lead Sarah Ryder) the animation quality is simply terrible. Any conversations about animations eventually drift into discussions about the uncanny valley. But what exactly is the uncanny valley and how does it affect animation, both in video games and otherwise?

Put simply, the uncanny valley is the idea that the closer that we get to photorealism, the more obvious mistakes become. The human eye is really good at picking out both human faces and things that are wrong. If two angles aren’t properly aligned, or if there are too many objects in a group, or if an object doesn’t fit. So when we combine these two traits, it becomes extremely obvious when mistakes are made in rendered characters. If lighting maps are wrong, or if a mouth is slightly out of sync with a voice, or if eyes don’t emote in the same way that the writing would indicate, it sticks out like a sore thumb.

These are all so difficult to do properly, and while many of the mistakes in Mass Effect are much more rudimentary (characters smiling when being told about dead parents), this is becoming more of a concern as graphics tech improves. Even in better made animation, this is something that we’ve run into for years. Look at 2004’s Polar Express. It was a great movie, but more than a few of the characters fell into the uncanny valley and it brought the illusion of immersion crashing down.

Anything less than absolutely perfect animation, and we will be able to notice every single detail that is wrong or out of place. As the years go on, we will notice these things more and more. The margin for error goes down the higher the fidelity of the animation, simply because there are more details that can be included, and therefore more opportunities for something to go wrong.

Then you have pre rendered animations versus procedurally generated animations, where you have so much less control over the end result. For a pre rendered scene, a company can spend millions of dollars to ensure that everything is perfectly rendered. There is a defined amount of time and you can just continue to polish the scene until it is perfect. But with procedurally generated animation, you have to build and map the systems and hope that everything works. You can’t control every camera angle or aspect of the environment or line of dialogue and how it will interact with every other factor of the scene. You have to trust that the systems you built work properly, and that nothing will screw up.

Sometimes they don’t, and we end up with something like Mass Effect: Andromeda. This is not to defend the animations here. There are amateur mistakes made all over the place that never should have reached the public, and it looks wrong in so many ways.

This leaves studios with a very interesting decision to make about how they will go forward. As tech gets better, “realistically” styled games will require more and more time and resources to stay ahead of this curve. It pushes studios towards more stylized art styles, which have the added benefit of allowing a game to stand out from its peers, while simultaneously avoiding the problems of the uncanny valley. With more stylized art, we don’t have the same confusion, and often the same level of detail is not required for it to look fantastic.

Where does that leave titles like Mass Effect: Andromeda, unfortunately caught in the middle, where it just looks bad. It’s stylized slightly, but just enough to make its characters look strange, while being unable to erase the glaring issues with the facial animations. If nothing else, it serves as a warning to other companies as to how they should treat their animations so they aren’t ridiculed in the same way.

_____________________________________________

You Might Also Like:

Tags: Film, Mass Effect, Polar Express, Uncanny Valley