Up until now, talking points in public conversations on bridging the fidelity gap between modeling and animation has been driven by marketing: touting talented actors and tech innovations. But even the most amazing actors and performances are captured in data: "In games, like film and life, it's how those individual bits of data come together that define a character's performance," says Mike Jungbluth, senior animator at The Elder Scrolls Online
developer Zenimax Online.
Jungbluth believes that a closer working relationship between animators, game designers and tool makers is key to creating believable, expressive animated characters in games. At Nordic Game in Malmo today, he spoke about how not only more collaboration among these disciplines, but also a stronger understanding of empathy, motivation and needs can make more lifelike characters in games.
"The traditional craft of animation is one many animators really hold onto as religion. But no matter how many people may want that to be the extent of a game animator's responsibility, that isn't the whole picture," he says. Tools have become increasingly invaluable and necessary as character complexity grows, giving more ownership to animators and less of the "toss the data over the fence conveyor-belt method," he says. More user-friendly tools means AI plays a bigger role in how characters communicate and express themselves.
"If the animators are visually-emoting, it's the AI brain firing off the synapses to play those emotions at the right time," Jungbluth notes. "To think either animation or AI can march out of the uncanny valley alone is foolhardy."
Animation and empathy
Game design builds the culture, rules and systems of the world that characters inhabit. "Those are powerful factors in informing a character's personality and performance. If AI is the genes that define the character's nature, the game design is the environment that nurtures the character," he poses. "If we want to keep pushing forward strong character performances in game, any of these disciplines alone isn't going to cut it. We naturally need to include and learn from each other, and make sure everyone responsible for a character performance is communicating the same personality."
"Games are not 'human', but we have to build in the illusion of an empathetic human character."
Even getting characters to move around the screen properly has long been an issue and can still be -- thus it's easy to understand why character performances have often been lacking. "Worrying about the larger performance can feel like putting the cart before the horse," he admits. "Getting believable movement is key in animation, but we need to move past this."
"What do we do with the characters once they're already moving and locomoting properly? And what do we do, beyond creating these individual animations that emote properly? Because those are really only the first step in creating a performance," says Jungbluth. "A great performance is all about what a character is going to do with those single animated emotions."
Players react to emotions through empathy, Jungbluth says. Empathy has become such an enormous buzzword in games in recent years, but breaking it down and understanding uniquely-human responses to the emotions of other humans are key to progressing the way players experience and identify with game characters. "Games are not 'human', but we have to build in the illusion of an empathetic human character," he notes.
"We can't create empathy in games the same way it's done in other media," says Jungbluth. In order for empathy to exist, distance must be present, he explains, a challenging wrinkle for games: "A person watching a movie has the component of distance built in as they're simply an observer, but a person playing a game is part of the action, often controlling the characters. This means crafting empathy can't be done in the same way for a non-playable character and a playable character. In giving control of a character to a player, distance is instantly compromised."
"Getting believable movement is key in animation, but we need to move past this."
Non-player characters need to have objectives -- often they either want to help or impede the player, and that can be all animators have to go on, too broad often even for design purposes. Traditional character lineups don't go far enough, he suggests.
Jungbluth shared a method for understanding non-player characters, like enemies, in relation to one another in order to understand and develop believable behavior for each kind. In Monolith's Lord of the Rings
game, he grouped enemies by intelligence according to their needs, whether a creature's needs were purely physical or more advanced. Each creature could exist on a spectrum: A warg might scoff at taking orders from an ordinary cave troll, but would be likely to respect a troll captain.
Goblins are more communal and have basic sociological needs, but are not much smarter than trolls, therefore one could expect them also to take instruction from a captain figure. Slightly more complex, orcs desire respect, esteem and achievements, but have low intelligence, and elves, high in intelligence and organization, would be on the opposite end of the proposed spectrum. This method let his team understand character behavior according to where they stood on a spectrum of needs, desires and sociological complexity.
Create an axis with binary traits on either end, Jungbluth suggests: Lord of the Rings
had a graph with "wild" on one end and "civil" on the other, and "sly" on one end of the intervening axis and "inept" on the other. Another coordinate graph presented "dominant" and "submissive" juxtaposed against "evil" and "good." A warg is wild and slightly sly, and also evil and slightly dominant. Plotting each character on this graph, especially relative to the player character, helps creators see where openings exist for needed personality and prevents groups from being too similar.
"In a case where characters are inherently similar to others in one way, you can also compare one character against the rest very quickly: Place a character's dot in the center of a graph, and then plot the other character's dots in relation to it -- instantly, you can see how they would behave if in a scene together, and how the social hierarchies would be established and played out."
By this point, you learn the character's objective, and how their life in the world is geared toward them achieving it, he suggests. But their pathfinding also is crucial to informing personality alongside AI behavior. For example, imagine skeletons conceived as "robotic guardians" that travel in uninterrupted lines. In contrast, a zombie's path would account for their meandering, lurching and stumbling, and is a messy, clumsy affair.
"These are behaviors we built into all the animations already, but if a zombie is given a path like a skeleton's, no matter how successful the animation, it's going to be at odds with how they move within the game."
With sociological charts and pathfinding plotted unique to each character, there's even more insight on how a character lives within a world and how they act within their ecosystem. Applying this specifically to animation is part of the task, but how to communicate each personality element to the team?
Communicating personality to the team
Level designers and world builders are in control of "staging" the characters using their specific animations, often having to choose from large and shared databases. Therefore it's important they're able to see the world in the same ways trained animators do. Books and documentations can provide strong and specific staging references, but just like when defining personality, a wall of text isn't sufficient and will rarely be returned to.
Jungbluth suggests a more specific approach: At Zenimax Online he was part of a group tasked with improving and defining staging areas within The Elder Scrolls Online
. He was able to give feedback on staging based on single screenshots -- often, where someone is standing in a scene communicates something to the player about who is the most important or most necessary to the player in, for example, a mage's guild. Visibility and posture within a scene can tell the player a lot about how they should feel about the characters they can see.
"Every character should exist before and after the scene they're in."
What does a character want? What actions would reflect that desire? Which emotions are they feeling? These, he suggests, are three questions animators can pose to designers that help them define the mood of a scene and how characters fit together. In staging, designers can identify the star of the scene, who they expect to help or hurt them in a given environment, and how they feel about the player -- is the player expected? What were they doing before the player arrives, and what do they want to do after the player leaves?
"Every character should exist before and after the scene they're in," he says. Attaching appropriate animations to these behavioral hooks help establish the character: It's not about adding more animations than are needed, but about placing them correctly. Behaviors like turns, starts and stops are directly informed by pathing charts, and each one should have a reason related to the character's objective. When characters stop, turn and move, it indicates a change of purpose for a character.
State changes can be powerful -- for example, the way BioShock
's Big Daddies can change state depending on what the player does enforces their objective. In Skyward Sword
, a change in state for a shopkeeper (he becomes dejected) when Link walks away from him deeply humanizes the character. Characters that can react to sound, smell or sight can be strongly differentiated: If one person swats a bug fervently and another does it languidly, you learn something about each of them and what makes them different.
Tying animation to player mastery
Unlocking new items are "performance paydirt" for animation: If a certain ability is always unlocked at a certain time in the game, designers should build that into the animation. If a character gets a grappling hook, for example, that should change their body language. They have the opportunity to react to being given a new item and change as they hold it. Progression systems are also powerful in that regard; those systems let players choose new mechanics and create actual growth that can be tied to a character's performance.
"Knowing when a new mechanic is unlocked allows animators to craft the motion and performance to react accordingly," he says. "Whatever their arc is, tying their animations to growth, to the player's mastery is really a no-brainer."
"When armed with what makes these characters tick, there's a great tool inherent to the medium that we can use to define mechanics," he suggests. Roleplaying as creatures, stepping into their shoes and thinking about living out their personality actually can help solve mechanical limitations and create ideas for what abilities or play styles might be lacking in playable characters.
Communicating is key to making everything work, he enforces. Everyone needs to be on the same page from the beginning to ensure their collaboration is viable within the context of scope, budget and time frame. But if animators, AI behaviorists and level designers all work together to define how characters express their objectives and need within the game world, the result will benefit everyone.