Developing virtual actors inspired by real actors’ practice

Abstract

Autonomous Digital Actors represent the next step in animating 3D characters. How to create such virtual actors remains an open question. This work presents and discusses an agent architecture derived from how real actors learn their art from acting schools. The knowledge base relies on a rule-based system where each rule encodes explicit knowledge into a classification vector which allows agents to suggest performances even for unforeseen situations, and to adjust their suggestions to match the animator’s preferences