MambaTalk: Efficient Holistic Gesture Synthesis with Selective State Space Models

NeurlPS 2024

Zunnan Xu,  Yukang Lin,  Haonan Han,  Sicheng Yang,  Ronghui Li,  Yachao Zhang,  Xiu Li 
SIGS, Tsinghua University
Paper arXiv Code
empty empty


MambaTalk is a gesture synthesis framework that leverages selective state space models to efficiently generate high-quality, diverse, and rhythmically synchronized 3D gestures and facial expressions from speech signals, addressing the challenge of high computational complexity.

Demo

BibTeX

@article{xu2024mambatalk,
    title={Mambatalk: Efficient holistic gesture synthesis with selective state space models},
    author={Xu, Zunnan and Lin, Yukang and Han, Haonan and Yang, Sicheng and Li, Ronghui and Zhang, Yachao and Li, Xiu},
    journal={arXiv preprint arXiv:2403.09471},
    year={2024}
}