Procedural Animation
polycounter
The programmer for Overgrowth gave a great talk on creating procedural animation, and the way he does it, it's a remarkably simple way of doing it that gives you a huge amount of control.
http://www.gdcvault.com/play/1020583/Animation-Bootcamp-An-Indie-Approach
I just wanted to share that. Any thoughts on the future of procedural animation in games? I know this was brought up in the E3 thread, and I felt we ought to discuss it more in depth. Any approaches you'd like to see attempted? Any examples of what hasn't been working so well? Got an interesting concept you want to prototype? Let's talk about it!
http://www.gdcvault.com/play/1020583/Animation-Bootcamp-An-Indie-Approach
I just wanted to share that. Any thoughts on the future of procedural animation in games? I know this was brought up in the E3 thread, and I felt we ought to discuss it more in depth. Any approaches you'd like to see attempted? Any examples of what hasn't been working so well? Got an interesting concept you want to prototype? Let's talk about it!
Replies
Procedural tends to seem to blend well with physics as well so it would be cool to see more impact with attacks and whatnot, if you get hit the stomach hard the body curve around the impact area before going flying.
basically I want to see it be used to have characters look more connected to their environment.
http://www.blender.org/wp-content/uploads/2013/11/2.71-fcurve_easing_equations_detail.png
Yeah, I want to see more connection with the environment and other characters too. I think everything he talked about in his video any game programmer could take and refine.
Pancakes: Yeah, that's a spring interpolation. That's the sort of animation curve he talked about using often.
The animations in this game are really something. some developers seriously need to take a leaf out of their book.
cant wait for more from them!
This is great for independent development. Definitely should be shared more.
13 frames for all his navigation is madness..
It's really interesting to see how the workload was reduced with his procedural techniques and to get a glimpse into the process of an obviously talented developer.
What I am thinking though, I might be wrong of course, is that in many cases there should be more of a halfway-meeting between code-driven, procedural systems and fully animated motion. While the procedural system does create a more linear workflow, maybe add more responsiveness, reduces the workload and makes everything easier to organize and handle and so on, which is really useful, especially for indie companies, who don't have a huge budget or giant teams, it feels as if too much control could be taken away from the animations if a game relies almost solely on algorithms and constraints for the movement.
Fully animated performances, can add a lot more to a character than the necessary locomotion and plausibility. It can add a certain and/or unique style, a certain personality to a character for instance. Look at sunset overdrive's amazing animation style for example, I believe a lot of that would be lost in a purely procedural system.
But I am generalizing here, of course it really depends on what kind of game you want to make, team size, the budget, what kind of style it has etc.
Most 3rd person games don't have great responsive movement, and the characters look like they are drunkenly bumping into things. Elder Scroll games are particularly bad in 3rd person.
That's a bit cynical, you just have to look closer to find some!
Jokes aside, there are multiple different factors influencing how something is animated and then tied into the game. Factors like performance, time, the genre of the game, the very nature of the medium that doesn't let you use movie pipelines, since you don't animate a "scene" but a specific move-set etc.
But I quickly did some youtubing and IF you are interested, you can check them out.
2) Wildstar has a small but cheeky creature race called "Chua". Take a look at their move-set, how the body and ears first squash when the character jumps and then stretch when the characters starts to fall, the way the body bounces back etc. The style fits the character perfectly imo. That can only be achieved when you have the control I mentioned in my other post.
3) inFamous: Second Son's combat animations. It has very clearly defined poses and for example the way the character's anticipation (the energy gathering phase) holds for a bit, while the chain unwraps itself due to it's momentum before he really rapidly swings the chain to hit the enemy for the first time (around 2:45), that is a cool way to convey strength and weight for example.
4) Witcher 2. It has really nicely choreographed combat animations. Sure they are mocaped, but they are still edited and retimed, some parts, I'm guessing, are completely done from scratch, here in some moves they hold the overall motion of the body while the arms continue to move forward etc. That's done make the moves look more powerful.
5) Batman: Arkham City's combat. Also nicely choreographed, also mocap. But the thing is this, it seems as though mocap is just the base for the motion, they took the mocap data and exaggerated a lot of the motions to give them a more 'comicy' feel to them, while the real world physics rules still apply. Realistic physics rules as a base while still exaggerating specific things for the comicy feel. You can see though that sometimes Batman will hold a bit, for example when he jumps in combat. That is a game design choice to ensure that the player has enough reaction time to continue or start a combo. Or a bigger enemy might have a long anticipation move before he hits, that's, again, to give the player enough time to roll away or counter or whatever.
Yes I agree, I don't like it either when there is no interaction with the environment etc. But that's why I said that ideally these two workflows would meet in the middle. Responsive movement, characters with more awareness of their surroundings but without sacrificing the control over the animations.
I don't really think that anybody hear disagrees that something might be lost from completely procedural animation, but what is nice about the system that David showed was that a lot of the animation in Overgrowth is controlled by keyed poses. So with that you can get a lot of your personality in the characters still while keeping the character quite procedural.
Not to mention the bounciness of the Chua in Wildstar is pretty straightforward squash and stretch which very well could be implemented into a procedural system. The end squash/bounce of the characters could be a more extreme example of the spring damper system of Overgrowths crouching.
Procedural animation doesn't mean that characters need to act like robots. Based on the way you go with it you can have characters show more personality and be more dynamic, It's just a matter of setting up your system to account for the types of body language and methods of environmental interaction that are required from characters in your game.
You can also see where he talks about improving procedural animation for more natural movements and more performance like animations in the Refinement and Ragdoll sections of his talk too. And not just that, You could probably even use the exact same system as Overgrowth but add more frames for variations in the character based on what they are holding, if they got hit in a certain limb and if you only have 13 frames for a character, it could make creating variations for different characters much easier as well.
Regarding the Chua, I took the jump just as an example, but yeah now that you've said it, if you use the spring damper system on the body and maybe offset the sections (stomach area, upper body, head etc.) you could create procedural overlap for the landing after the jump and by adding drag to appendages like ears and tails, combined with key poses you could get the same effect procedurally.
I mean, don't get me wrong, I could definitely see the advantages of Dave's workflow, but I got a bit caught up in the idea/worry that code-driven, mathematically approached motion = loss of an artistic touch. But I see now that I missed part of the point and that it depends on how you go at it and that you can expand that workflow, like you said. Thank you for explaining, Shadownami92.
If you think game animation is bad and unpolished, only through the intersection of code & animation will it get better.
ANYWAY I'm absolutely going to steal this technique for my games, I've been really interested in procedural content generation for the last 20 or so years.
Through the talk he made it clear that the system is developing as it goes along. Constant iteration of this system will fuel the look thus raising the quality. (if you believe that the animation is subpar)
It's also worth mentioning that I believe the majority of these cycles, as he said in the talk, are running from either 3 or 4 frames, maybe even less in some scenarios. Product management triangle comes to mind, and I'd argue that Wolfire is heading in the correct direction. Hats off to the dudes there, this was a really inspiring talk for me as an animator and more importantly, a developer.
I would do prefer it if more developpers would focus having quality animation like these..:)
No I don't have anything against the game animators. It's more about how they're trying to cheapen and cheapen animation cost in video games resulting generally with characters that have soulless automated motions.
I don't know what Sunset Overdrive is, but Witcher 2's responsiveness is similiar to Dark Souls. It's "clunky" on purpose in order to simulate the weight and follow through of a hero wrestling with inertia.
Sort of like how the protagists in Castlevania games have the delay before and after each attack in order to simulate the fact that, "hey you're fighting with a whip, man"
Sometimes it's a relavant design choice to lower responsiveness.
ANother example would be the survival horror genre, you can't give someone the ability to double jump and parkour when they are supposed to be afraid of the zombies that are inching toward them slowly through the streets. Well maybe you could, cuz that might be sort of awesome, but it might diminish how scary it was if you did that.
[ame="
One could easily and correctly make the argument that it's soulless to have animation pre-made that can't possibly take into considerations all the variables that happen in a game world.
@Blond: Your avatar is done well, insofar as it is very expressive.
But (and maybe you're fully aware of this): People will read everything you have to say and match it at least subconsciously with this expression in front of them and it doesn't help that both seem to be a perfect match not too seldom.
As for the topic: I think it's very clever and looks better than I would have expected with that few keyframes, though a quick moving character like the one in the demonstration might lend itself more to this technique than a slower one.
Which btw is totally relevant to procedural animation.
[ame="
Or maybe this guy.
I seriously hate when people laugh at stuff like this..
what's the joke?
If I was the guy presenting this I'd just walk off...
oh a cube is walking!! HAHAHHAHAHAH OH LAAAWD THAT SURE IS FUNNEH
that might add to the subject.
I'd say I was refering to the crowd in the video and not to you though but sure whatevs
The things like secondary motion on the ears are hard to do when you have 3 animations blending together. How do you animate a scabbard reacting to being hit by 2 forces at once when they don't know when or how much force will be applied? In film you know exactly when and what is going to happen, you can plan for it and account but in games a lot of that is controlled by chaos.
If it's handled with procedural motion, it can work.
If you have an animator do it, they waste their time doing a bunch of "what if" animations making a giant bloated animation tree. Ask any game animator and they'll probably say they hate that part of their job because it breaks easily and is one of the more limiting parts of the job.
It's not about doing it cheaper, its about doing better animation that responds to the unknown forces that are outside of animators control. I understand the whole "John Henry vs The Machine" thing and I don't like studios being lazy about their animation but I do like people being very smart and efficient with their animation and actually coming up with ways to handle those unknown forces.
He's being incredibly efficient and is able to deliver higher quality, more responsive animation with less work (on the animators end). Where other studios throw more and more resources hoping the problem magically disappears into the sea of faces.
That doesn't lead to higher wages or higher moral, that leads to a lot of low wage labor that ends up hating their grunt jobs because it's all a bunch of grunt work. At best they can aspire to be the lead or manger that oversees all those grunts, which just really means they are a "bitch filter" so upper management doesn't have to listen to the whining.
How he approaches animation is VERY much how traditional animators (Les Clark, Milt Kahl, Eric Larson, Richard Williams ect...) would animate. Create key frames (the major frames that define the character and poses). Once the major motion flows right, let the assistants connect the dots. In this case his code handles the in-betweens so the animator covers more ground and focuses on what is important, defining the heart and soul of the character.
Brilliant.
It's how animation has worked for a very long time. Well... except for the ideological purists who force their coders to use punch cards and giant tape reels and force their texture artists to paint in oils and then scan them in once they dry. "That's what real ______ do!" BS, heh.
Not to mention he is probably saving on a lot of technical overhead which means they can pour those resources into something else.