Researchers at the University of Toronto (U of T) in Canada and the University of Massachusetts Amherst have used deep learning algorithms to improve software for the animation and gaming industry.
The team tapped insights from psycho-linguistics literature to produce VisemeNet, a network for predicting visemes (the appearance of mouth shapes), which map to vocal cord sounds.
They blended the results of phonemes and facial movements, using an actor's voice as audio input to predict speech motion curves that are fully editable in animator software.
VisemeNet was developed as a component of jaw and lip integration, which enables animators to create realistic and expressive speech animations of computer-generated characters.
From U of T Engineering News
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA