Sleepy Brain: Orc Army

by Miles Merrill
Published in association with The Program

Sleepy Brain: Orc Army

“Battalions of Orcs of the Eye and countless companies of men of a new sort that we have not met before… The numbers…could not be guessed in the darkness…fear…had scarcely overcounted…dark with their marching companies…Busy as ants, hurrying orcs were digging…just out of bowshot.”
– JRR Tolkien, Return of the King: The Siege of Gondor

The Skeleton of a Digital Agent
Peter Jackson’s The Return of the King features a spectacle of extraordinary battle scenes. Big, Huge…Massive. Hosts of Mordor, Riders of Rohan, Men of Gondor and Shadows of the Dead in “six figures” according to Stephen Regelous, inventor of the ground-breaking software Massive. After working together on the poltergeist film The Frighteners (1996), Peter Jackson approached Stephen Regelous, digital artist for Weta Digital FX New Zealand, about designing software to create realistic crowds for his new “Hobbit” project. The result was the now famous software Massive. Get Ready. Massive will soon be available to artists in the film industry as a 7MB downloadable application.

Actions
Using a library of more than 300 large and subtle movements derived from motion-capture sessions with over 350 real costumed actors, Massive allows an animator to build a realistic looking character with specific actions and reactions to their environment – run, strike, block, walk, charge, turn, etc. They don’t have to follow a predetermined cause and effect pattern unless specifically programmed to do so. An example could be: if an animator designs 12 characters, each with different physiques (skin, clothes, weapons, tall, resistance to injury etc.) each having thousands of different reactions, he can duplicate these 12 and use input/output nodes to populate a shot or scene with unique variations of the originals. Literally through 12 characters, the animator can create millions of similar bodies all performing different tasks at different moments within a sequence like the Battle Of Gondor in Return of the King.

What the Massive animator sees is an interface resembling a spider-web or neural network connecting actions with figures on a digital plane. The ideas come from fuzzy logic experiments and iterative (self-repeating natural pattern) algorithms in engineering. Instead of agents making binary- yes/no, true/false- decisions they are capable of using details around them to make subtle and natural choices allowing them to avoid that rock, climb that hill, whatever their digital environment requires. All of these real-time functions are controlled primarily by psuedo-senses: “hearing, sight, and touch”. These are basically information gatherers that navigate through environmental changes. “By simulating the way individuals react to their environment and each other,” explains Regelous, “we got realistic crowd behaviour for free.”

SENSES: Hearing
Each agent emits a specifically pitched sound which marks their group- human, Orc, car, horse… and can respond to the noises made by other agents. Using a flocking-type command for example each can be directed to avoid collision with other sounds. This assists particularly in group movement.

Sight
Characters have a rendered visual image of their digital environment for every frame of film. So their brain (software) never has to scan every piece of data on the map (database). They can ignore an enemy two miles away and concentrate on obstacles and other agents in front of them at 24 frames per second.

Touch
Agents are provided with a means to detect collisions which relates with what they run into. So they can have a set of responses for terrain- rocks, ponds etc. and more complex responses for colliding with other agents.

When all these factors are combined, the filmmaker can break scenes down, programming tendencies to make sure for example, the good guys win. Agents that run from a battle for no particular reason can be adjusted within seconds for another take. Movements and responses can be manually tweaked but if you’re looking for reality it’s more fun to watch the little figures act out their predetermined destinies.

I asked Regelous: Does he believe an AI character will turn to him one-day and demand: “What are you looking at”?

He said, “Sure, I think we may one day have agents that don’t realise they’re computerised. For example, a lizard that really thinks it’s a lizard. One of the questions I have is when we start programming virtual stuntmen to feel pain – because that’s certainly coming – who’s responsible for their well-being?”

Never have film effects held such powerful philosophical implications. The medium is set to eclipse the message. One day a film like Spielberg’s AI will be a hand-held guerilla documentary.

– This is an excerpt from Miles Merrill’s article, Where’s Massive?, to be published in Metro Magazine’s upcoming issue #139