Grand Text Auto

This past January, Craig Reynolds from Sony’s research group in Foster City gave a presentation at UC Santa Cruz on current challenges in creating computational crowds, especially those where the members of the crowd are cooperating to perform some task. A video of the talk can be viewed here:

https://slugtube.soe.ucsc.edu/play-video.php?ID=1072

As a contrast to the recent talk by Mark Henne from Pixar, this talk focused on the underlying algorithmic difficulty of creating the desired algorithmic behavior. Mark’s talk focused more on how to integrate behavior into an existing pipeline, and challenges with ensuring the filmmakers retained artistic control over the procedurally generated scene.

On Friday, May 29, 2009, Mark Henne from Pixar came to UCSC and gave a talk on crowds in the movie Wall-E. You can watch the video here:

https://slugtube.soe.ucsc.edu/play-video.php?ID=1070

I was impressed by the complexity of the AI underlying the characters that comprise the crowds, another reminder of just how complex seemingly simple real-world behaviors can be. I was also struck by how the entire process was optimized to ensure that the artists could, if desired, take a single character from a crowd and manually change its look and behavior. This is consistent with the entire filmmaking process at Pixar, which is optimized for complete artistic control over the end product. Games, in contrast, are much more willing to accept the limitations of the game engine being used.