Courses

FCFull Conference Pass (FC - All Days)
FC1Full Conference Pass (FC - 1-Day Only)
BCBasic Conference Pass
ExExhibits Only
ETElectronic Theater Ticket
RTReception Ticket
Title
AI in Feature Film Animation: How Crowds Artists use AI Techniques at Pixar
Date
Wednesday, 04 November
Time
09:00 - 12:45
Location
Kobe Int’l Conference Center, Room 504+505

AI in Feature Film Animation: How Crowds Artists use AI Techniques at Pixar

Courses Information

For much academic research as well as industrial applications, the goal of crowd simulation is to be maximally predictive. In animated feature films this is not even a secondary criteria: a crowd simulation is judged successful to the extent it is believable to the viewer, and directable by the creator. High levels of believability (we won’t say realism, the presenters work more often with talking animals than humans), is the baseline standard for professional work, but ultimately, the crowd simulation must serve a specific story point. That is where directability comes in, a crowd simulation for film must be amenable to nuanced crafting to match a director’s vision. This can often be in direct contradiction to believability, which is where the skill of the crowds artists and the features of the simulator are put to the test. These skills and features will be the focus of this course. We will start with the use of Finite State Machines for animation clip sequencing, and show applications in “A Bug’s Life”, “Cars”, and “Up” for stationary crowds. For moving crowds, we will discuss Agent Based Simulation, how Pixar integrated Massive into its crowd pipeline, and built powerful and flexible brains for agents in “Ratatouille”, “Wall-E”, “Up”, and “Cars 2”. Multi-source navigation, signal processing for physics, subsumption architecture, and time till impact collision avoidance are among the techniques discussed. Next, we will take a step back to Finite State Machine sequencing and present a more sophisticated pipeline and user interface for crowd simulation on “Brave” and “Monster’s University”. Finally, the course will conclude with Pixar’s new agent based simulation framework, MURE, and explore how Finite State Machines, Fuzzy Logic, Rigid Body Simulation, and fast geometry cache sequencing can be implemented in a Houdini FX pipeline.


Paul Kanyuk, PIxar Animation Studios
Stephen Gustafson, PIxar Animation Studios

Paul Kanyuk is a Lead Technical Director at Pixar with credits on Cars, Ratatouille, Wall-E, Up, Cars 2, Brave, and Monster’s University. His specialty is crowd simulation, shading, and rendering, and is responsible for the procedural animation and rendering of numerous crowd spectacles, including the hordes of rats in Ratatouille, deluge of falling passengers in Wall-E, and pack of talking dogs in Up.

Stephen Gustafson is a Technical Director at Pixar with credits on Up, Cars 2, and Brave. He is a computer graphics wizard, and the primary architect behind Pixar’s proprietary crowd simulation framework, MURE.

Intended-Audience:
Computers graphics practitioners of all stripes from animators to technical directors are welcome. In particular, we’re hoping to draw attention to the aspects of crowd simulation and AI most important to feature film animation, and hope those interested in learning about this topic, even from other fields, can attend.

Go Back
BACK TO TOP