MACHINE LEARNING
We utilized MIDJOURNEY AI to generate hundreds of destroyed camp props for scatter in our environment; toppled tents, blood covered weapons, armour, broken wood and twisted metal. Using carefully crafted prompts we circumnavigated the list of banned words and were even able to generate alpha extracts from our output for use in DMP and COMP departments respectively. This strategy saved us thousands of design hours as we’re able to generate thousands of variations in seconds.
TREBUCHETS:
We streamlined our trebuchet animation by standardizing the launch cycles of each trebuchet which showed the loading, launch, and reloading cycle of equipment and characters involved in the action. This included their motion captured performances and their cloth simulations along with a dynamic release of each bomb. Each animator/layout artist was able to easily offset the trebuchet animation to match the timing of photography and an evolving edit without reopening animation tasks.
FX SIMULATIONS
The CGFX team was tasked with massive explosions which needed to destroy the camp and everyone in it. These complexities under tight production schedules meant we need to innovate!
We prototyped and implemented parallel procedural workflows where multiple artists were able to work in real-time on individual components of the same effects in the shot. Task automation was used to create a co-dependent chain of events to be executed and monitored on a higher level. These simplified and non-technical destruction toolsets allowed the artists to focus on the creative aspects of shot production.
The quality expectations for the show were set extremely high. This meant very long heavy simulations which amounted to terabytes sized cached data which needed systematic clean-up, LODs and proxies, and extensive TD and IT support.
Render time was extremely long due to the decision to render the hero shots with the highest quality settings and full GI. A factor that greatly contributed to the final look of the finished shots!