PersonaGest: Personalized Co-Speech Gesture Generation with Semantic-Guided Hierarchical Motion Representation
The PersonaGest framework represents a significant advancement in the realm of co-speech gesture generation, particularly for graphics programmers and animators. By utilizing a semantic-guided RVQ-VAE and a Masked Generative Transformer, it effectively disentangles motion content from gestural style, allowing for more personalized and contextually relevant animations.
This innovation not only enhances the quality of character animations but also streamlines the workflow for developers, making it easier to create expressive and coherent movements that align with speech. As the demand for realistic interactions in games grows, tools like PersonaGest could become essential in the toolkit of any developer focused on immersive storytelling.
“Extensive experiments demonstrate state-of-the-art performance.”
- what
- Introduction of PersonaGest for personalized co-speech gesture generation
- who
- Developed by Junchuan Zhao, Qifan Liang, and Ye Wang
- when
- Submitted on 8 May 2026
- impact
- Improves realism and personalization in character animations
The framework offers significant improvements for animation realism.
Follow gesture generation updates
See relevant stories in your personalized news feed.
Discussion