GATHERING

TitleGATHERING
BrandMEDIA SCAPE
Product / ServiceCIGNAL
CategoryA02. Computer Graphics/Visual Effects
EntrantFILM PABRIKA Makati City, THE PHILIPPINES
Entrant Company FILM PABRIKA Makati City, THE PHILIPPINES
Advertising Agency DM9 JAYMESYFU Makati City, THE PHILIPPINES
Production Company FILM PABRIKA Makati City, THE PHILIPPINES

Credits

Name Position
Merlee Jayme Chief Creative Officer
Louie Sotto Executive Creative Director
Aste Gutierrez Copywriter
Biboy Royong Art Director
Jem Lim/Alec Humphries Agency Producer
Marivic San Juan Producer
Stephen Ngo Director
Central Digital Lab. Special Effects/Computer Graphics

Brief Explanation

Half a million people are shown rushing somewhere only to reveal that they are rushing home to enjoy their favorite TV show in HD clarity, thanks to Cignal Digital TV Broadcast.

Creative Execution

CROWDS We considered a couple of systems currently available inside 3dsmax. One was to use Particle Flow to create cached animated geometry of crowds. The system also utilized controlling each particle crowds’ age to offset the animation. But this later proved to be too taxing to control and handle due mainly to the large amount of instanced objects and some work making convincing crowd flows. Another system we tried out was the old Crowd System that came with the Biped plugin. After implementing some custom scripts and low level Motion Flow graphs to control crowd behavior, the initial tests and results had some potential. But we later decided to scrap the approach as the system was not very much suited to handle larger crowds in the amount of time we needed - where it mainly needs a good amount of animation library. Another big factor was that both PFlow and the Crowd System was handling simulations using only a single core of our i7 dual quad core rig. In the end, we went back to a simpler approach, coupled with some newer tech in handling cached animated geos which resulted in two complementary systems. A. For shots needing directable crowds, we animated a single run animation using the live footage as reference were it was later point cached and looped. The animation was then transferred to 7 unique mid-low resolution character models, each was path deformed to a spline to control the running direction. Custom attributes were also created to offset the animation and control the run speed. The computation of the attribute automatically adjusted the animation based on the user preferred settings. B. For larger crowd groups, we utilized building preset groups from the spline system (A). We created random clumps, lines and directional crowds that were all cached out as one and used to fill in most of the empty areas in the shots. Intersecting groups didn't pose much of a problem as you could barely notice it from the running chaos on farther shots. C. We also used Softimage XSI ICE crowd system for the thousands of people for the really wide aerial shots. We manage to create 1000 to 5000 CGI people with random goals and motion to make them believable. Lastly, we developed a tool to massively control the crowd animation offsets and material randomization to further create a more natural and arbitrary feel for each shot.