AI Characters from Animation to Behaviour


Christiaan Moleman (host), Xavier Dolci , Phil Carlisle, Julien Hamaide

This panel brings together an artist/animator, and animation programmers/designers to discuss the challenges of next-gen animation. How can we improve our current animation workflows to get around the usual motion capture problems? How can we improve the quality of characters without having gigabytes of motion capture?

Hopefully I got the names matched up with the people correctly, this was a great set of questions and in interesting look at the easy and harder ways to improve things in general, and also why it should be improved!

C: What can animation convey which would help AI? Any projects you’ve worked on that work on this?
J: At a time the player has money powers, animations used to convey the monkeyness which wasn’t realistic but conveyed why they could jump high, but it is close to the story and main AI concepts. That is why it is difficult to work with animators who want to do the most realistic thing possible.
P: Absolutely correct, don’t need realism. Look at Disney, hyper-realistic. Intentionally use boxes as art for projects for this.

C: On realism, The Muppets are pointed to as an example because of the limitations of their medium, so it’s movement, gaze direction and timing!
P: Never ask a programmer, as artists say. “Everything is too slow” they complain. Looking at puppets though, and applying the techniques to proper faces would be a good idea.

C: Also like South Park is a lot more expressive then game characters. Gaze direction and eyes seem huge. You’re playing with zombies – not all games have zombies in them though! People should look at you in the eyes, like Half Life and Valve.
P: Bottom eyelids – look at Disney examples which have moving bottom eyelids.

C: A real smile versus a fake smile have the muscles scrunch up, while a fake smile is just the mouth moving. That has a lot of effect making them believable, and is a huge difference. In AI, if you have a character communicating what they’re going to do they look at it.
X: Should have the animators and programmers investigate the psychology and sociology and so on.
J: Graphics programmer knows how to do lighting since it is just mathematics, but knowing what to show is difficult. We don’t at the situation with scientific eyes. As a programmer you take everything like you know how to do it.
P: It’s a huge amount of observation. Filming people, in secret and invoking them – hard for researchers, and while it is dodgy it is observation. Find the most expressive person in your studio and annoy them! Fantastic stuff.

C: Any examples of games that do this well?
P: I gave some examples yesterday, but I’m not convinced by many of them.
J: The examples on the slide are old, but it does take work to do facial animation, but it adds so much to it.
P: It’s totally not technology, we can do all these thing. Valve also got asked to do a film with TF2 since the trailers were so well received, so it can add value like that too.
X: You have to fight to put it in the game, try it and see it.
J: People are saying “want the best lighting” or “best graphics” but never “best animation” although when the systems are released they have huge success. Need to change minds in the industry.
C: There’s a perception in animators in industry that they’re just doing walk cycles and death animations, and no way to get this added. Need to get in early and have the animation team work on this, with a cross discipline team, from all sides.

Question: So going from the muppets, but the games with detailed characters need very good animations to match, with a small budget how can they cope?
C: The thing to take from the Muppets is not to do puppet animation but to get your priorities straight – get the Muppets stuff in there first, gaze direction, looking, timing. Important for the thought process in a character. Then move on from that to get more subtle expressions and animations.
X: Are in the 3rd generation of motion capture, can capture the body but not the face. Fighting that but not getting good results.
J: Industry is spending much on graphics engines but not animations. There are a few like NaturalMotion, but having more would remove a big weight from the animators if they had those systems. Need to find a solution from a programmer point of view.
P: Adding some proceduralness, not into everything but into a few of the animations like happy walking, slow walking etc.

Question: Question of expectations – a realistic looking characters having brilliant expressions, versus Wall-E or South Park having realistic faces and expressions. What can we do to improve this character, and two player immersion?
P: Many games a lot of the background stuff like hand gestures. Lose a lot of that in games.

Question: Industry for a reason – making money. Maybe it doesn’t sell, maybe customers doesn’t want it.
P: Could say the same thing about shaders.
Q asker: but it costs money.
P: Do incremental improvements.
Q asker: Depends if you get the project management to do it. And no one complains about it.
J: It really connects the player to the character. Everyone can emotionally connect to and for building a franchise could bring it out.
P: Point of differentiation between games.
C: Not just a point of immersion, but also a game play thing. Ico or Shadow of Colossus without good animation is a major part of the game – the level design even for SoC. Body movement is important, facial animation is icing on the cake. If you choose not to have that interaction fair enough, but it needs to be there for proper interactions.

Question: What about an SDK for doing what the animators do?
J: There are some SDK’s, but the problem is people do not like to buy or test middleware all the time, and have to tie one or two animators to the software, which they don’t like. It’s not technologically hard to implement, normal animation systems have most of it already. It’s always the same, have to convince the manager it works, learn it as a programmer, push it to animators and artists. Hard to push since they are right, learning new tools is a difficult thing.

Leave a Reply

Your email address will not be published. Required fields are marked *

Website and journal of Andrew Armstrong