Article
i/2008/07/willspore

Planting the Spores of Procedural Animation?

Dave Mark on July 24, 2008

In week’s developer discussion, Dave Mark looks into the subject of animation, whether procedural, parametric or physically-based. Let him know what your take is on this kind of technology and post a comment below.

If you think walking on 3 legs is hard, try brushing your teeth when your hand is hanging over your back!

It’s no secret that one of the more anticipated titles of the year will be Spore (Will Wright / Maxis / EA). With this title, the pre-release buzz has been almost peculiar. It’s not of the same genre as a typical AAA title like a GTA or Halo. It almost has a Molyneux-esqe mystique to it. No one quite knows what to expect out of it.

One of the reasons for this may be because what is going to be in it. This was certainly brought into focus by the release of the Creature Creator recently. It was at that point that people realized the full magnitude of what was going on. You were going to be able to putz with your creations to an unprecedented level. To quote Will, the Creature Creator is like combining Mr. Potato Head, an Erector Set, and clay… all of the relatively free-form putzing tools from my childhood. (Remember I’m really old!)

We aren’t just talking about changing your clothing or armor or hair color or the shape of your mustache… we are talking serious putzing. You can change the shape of the body - but not just limited to making it short, tall, thin or fat. We are talking about really changing it to things that are borderline ridiculous. Never mind how long the legs are… how about dealing with 3 legs… 5 legs… 17 legs… whatever you want. Never mind the color of the eyes… how many eyes do you want? Where do you want them? Arms? Well I suppose that depends on what you consider to be an “arm”, doesn’t it? Seriously… stuff gets whacked out beyond belief [surprisingly, this is not a technical term like “putz” is… not if my 16-year-old daughter uses it regularly in conversation.]

And all of that makes for a bit of a problem when it comes to animating the sorts of creatures that we have seen get churned out by the public. Pardon me for being presumptive, but I suspect we haven’t got a lot of MoCap data sitting around that would apply here. And yet, it works. And, according to Will’s presentation (Video 1 below), it’s all due to the magic of “procedural animation.”



Video 1: A 35-minute walk-through of Spore by Will Wright from 2005. While you watch, count how many times he says “procedurally”.

The Euphoria over Procedural Animation

Natural Motion’s version of the Michelin Man.

In AI circles, there has been no shortage of discussion in the past few years about procedural animation. I’m not an animation guy so I haven’t burned as many clock cycles of my own on it as I do with other stuff. Still, it’s kind of hard to miss. There has been plenty of attention paid, for example, to Natural Motion’s “Euphoria” engine — and especially with regards to the upcoming game Star Wars: The Force Unleashed. In March, Alex wrote an slightly controversial editorial column about it as well. Also, over the past few months, there has been an extensive thread on the AiGameDev.com forums entitled “Unlimited Procedural Animation” that has been home to various opinions, suggestion and speculation. (Free forum registration and introductory post required to view.) If you want some really heady reading from EA’s Chris Hecker (et al.), behold the paper “Real-time Motion Retargeting to Highly Varied User-Created Morphologies” where he talks about the methods involved in making the Spore monstrosities move.

Perhaps it is a good time to clear up definition of terms. According to Wikipedia, “procedural animation” is simply defined as…

“… a type of computer animation, used to automatically generate animation in real-time to allow for a more diverse series of actions than could otherwise be created using predefined animations.”

I’m not sure I even want to see this thing walk. It makes my back hurt just to look at!

The use of PA in Spore actually takes a twist on the above statement. I would contend that the above phrasing would need to be amended to say…

“… a type of computer animation, used to automatically generate animation in real-time to allow for animating relatively simple actions from a more diverse pool of potential actors whose potential construction renders predefined animations completely useless.”

In support of my restatement above, I submit Video 2 below. There are some amusing statistics about the Creature Creator’s success… but more importantly, gape at some of the creations that have spilled out of the collective mind of the masses. (If you really get bored, go search YouTube for Spore Creature Creator Video. I would suggest it is safe to assume that the overwhelming numbers of genitalia-based constructions won’t make it into the distributed content. In the mean time, surf with care.)

Video 2: Will Wright presenting the latest on Spore at the 2008 E3. He talks heavily about what the public has done with the creature creator.

It would seem that PA is really hitting the big market… and certainly the user community and the (semi-)objective media will let us know what they think about the results. However, do the successes of packages like Euphoria and the PA that is being utilized (by necessity) in Spore, give us something to chew on for the future of animation in the games industry? Some industry folks have offered that seemingly esoteric techniques such as this may not be able to be applied across the spectrum of games. I’m not so sure.

Pressured Animators vs. Procedural Animation

It is not secret that content creation is the major portion of development time and expense these days. Putting together a game engine is loose change compared to having legions of artists and animators working on every single character (main or background), every single behavior (spectacular or mundane) and every single game situation (unique or repetitive). That’s a lot of man-power and a lot of budget that is earmarked for this process — earmarks that we don’t necessarily have to burn. Let’s face it… for every blockbuster mega-seller that turns a profit, there are scores of games that don’t. If there is a way to help in streamlining the process, it not only translates into dollars, it could translate into projects that don’t get axed, studios that don’t close and lay-offs that don’t happen.

But is PA enough? Sure, the knock right now seems to be that it deals with specific case scenarios. On the other hand, so does MoCap data. If you want an accurate military crouch walk, you get the MoCap. If you want an accurate football pass animation, you get the MoCap. If you want an accurate nose-picking animation, you get the MoCap. However, what if you decide that you want something else? Of course you can always resort to animators hand-crafting things. Nose-picking, I suspect, would be simple (although I have no frame of reference for animating such things)… but a reasonable rendition of a Riverdance might be a fairly involved process. Be that as it may… do we really need to spend valuable time animating walking, running, jumping, falling, throwing, catching, blocking, pushing, shoving, being pushed, being shoved or any such perturbation?

First step… find the center of gravity!

I don’t know enough about the whole animation gig to opine one way or the other on the graphics and animation pipeline… but I believe that there is a valid question here. Even though PA may not be the silver bullet yet, is it something that can be massaged and tweaked to eventually fill an expensive role in content creation? If not, what else needs to happen to the field to make it worth our while? For that matter, if it can be use, what sorts of nifty things do you see it doing — aside from providing somewhat reasonable locomotion to the occasional mutant creature with a center of gravity that is in a different time zone than it’s head?

Discussion 3 Comments

alexjc on July 24th, 2008

[B]A)[/B] What they're doing in Spore I'd class more as "parametric animation" since it's much closer to the typical [URL="http://aigamedev.com/reviews/crysis-animation-integration"]motion graphs + blending techniques[/URL]. In fact, Lucas Kovar, who worked on parametric motion graphs with Mike Gleicher and Rachel Heck, is listed in the [I]Thanks[/I] section of the Spore paper. I call this parametric because it's heavily based on adapting animation data -- like Spore is. [B]B)[/B] Procedural animation, generally speaking, tends to be defined as [URL="http://www.mrl.nyu.edu/%7Eperlin/"]Ken Perlin[/URL] does it: every track and keyframe is animated by a procedural equation. So this approach is heavily based on math. [B]C)[/B] Euphoria uses a third technique which could be labeled physical animation. Basically the postures are computed each frame by running a simulation, in this case based heavily on physics. So, basically I don't think Spore is a good example of cheap animation that comes to the rescue of low budgets (it's probably the opposite). Perlin's procedural stuff is much mure suited to this, but it takes much more programmer time. As for euphoria, well... see my [URL="http://aigamedev.com/editorial/naturalmotion-euphoria"]editorial[/URL]. In general, if you're in AAA games and not based on A) parametric animations, you're doing something wrong. If you're an independent developer, go for B) procedural Perlin-style as much as possible. Alex

Kevin Dill on July 24th, 2008

These techniques are exceedingly cool - like as in I have to wipe the bits of drool out of my beard every time I go watch the movies again - but I don't think they're going to be putting animators out of work anytime this decade. The big advantage Spore has is that [B]you've never seen anything like those creatures before[/B]. This means that while they have to get something that looks cool and believable, they don't have to make it look *right.* They don't have to make it look *real.* You'd be amazed how quickly the eye will catch an animation that looks wrong for a human character or an animal we're familiar with. That's what the uncanny valley is all about, and the folks at Maxis have entirely dodged that issue in this case. That's not a criticism - both the concept and the execution are brilliant. But I don't think the technique generalizes. Natural Motion is also amazing, and what it does well it does extremely well, but what it does best is basically to make ragdoll look like it should look. When you try to move characters around with it in the absence of external influences (for instance, their demo with the character waving or saluting), the result looks very, well, ragdolly. If I can be permitted to make up such a word. It's not crisp, and again, it just doesn't look quite right.

Dave Mark on July 25th, 2008

You know... to [I]dork around[/I] with it. (Go look THAT up... I'm sure it won't be much better.)

If you'd like to add a comment or question on this page, simply log-in to the site. You can create an account from the sign-up page if necessary... It takes less than a minute!