The tradeoff for the gigantic film-quality results, however, was a render speed of 6 hours per 50 frames, leading to hundreds of hours of baking on a 15 machine AWS cluster. So, to transform this large film asset into a real-time game character, Ziva fed over 12GB of performance capture data along with the high-quality Ziva VFX simulation into their Ziva Real-Time Trainer. This technology used machine learning to train the troll asset to perform all of the animations along with novel poses in real-time while maintaining the rich dynamics of the original simulation. As a result, the final ML troll body performed at a fully-interactive frame rate of under 3 milliseconds per frame in Unreal Engine 4.26 and was ready to be handed back to the innovation-loving Ninja Theory team.

"By utilizing their latest runtime technology Ziva helped Ninja Theory bring the full fleshy horror of the Troll to life in a way conventional workflows just can't do." says Andrew Vidler, Technical Director at Ninja Theory. Since its inception, Ninja Theory has been devoted to pushing the envelope for interactive digital content. They have continued to work with new, paradigm-shifting technology teams to uncover and create the bleeding-edge of gameplay experiences, and the partnership between Ziva Dynamics and Ninja Theory is set to generate even more best-in-class character results into 2022.