Sebastian Starke

Sebastian Starke will present about the most recent advances in data-driven character animation and control using neural networks. Creating key-framed animations by hand is typically very time-consuming and requires a lot of artistic expertise. Instead, by learning from motion capture data, several works applying deep learning for character animation have demonstrated being able to produce high-quality motion controllers for complex tasks, such as biped and quadruped locomotion, object interactions or sports animations. Those works were found being able to compete with, or even outperform the quality that could be achieved by traditional techniques or professional animators, while heavily compressing the motion data and generalizing to novel situations. He will showcase a number of results using such AI-driven character controllers, present how such systems generally work, what makes them different from existing solutions, how they can integrate with current animator workflows, and which potential impact they may have on future technology for video game development.