Animating an Autonomous 3D Talking Avatar
One of the main challenges with embodying a conversational agent is annotating how and when motions can be played and composedtogether in real-time, without any visual artifact. The inherentproblem is to do so — for a large amount of motions — without introducing mistakes in the annotation. To our knowledge, there is noautomatic method that can process animations and automaticallylabel actions and compatibility between them. In practice, a statemachine, where clips are the actions, is created manually by settingconnections between the states with the timing parameters forthese connections. Authoring this state machine for a large amountof motions leads to a visual overflow, and increases the amount ofpossible mistakes. In consequence, conversational agent embodiments are left with little variations and quickly become repetitive.In this paper, we address this problem with a compact taxonomyof chit chat behaviors, that we can utilize to simplify and partially automate the graph authoring process. We measured the time required to label actions of an embodiment using our simple interface,compared to the standard state machine interface in Unreal Engine,and found that our approach is 7 times faster. We believe that ourlabeling approach could be a path to automated labeling: once a sub-set of motions are labeled (using our interface), we could learna prediction that could attribute a label to new clips — allowing toreally scale up virtual agent embodiments.
Computer Graphics, Visualization, Computer Vision and Image Processing (CGVCVIP), 2019 (Best Paper)
Dominik Borer, Dominic Lutz, Martin Guay, Robert W. Sumner
Paper (coming soon) — Video (coming soon) — Iadis Digital Library