TANGO: Co-Speech Gesture Video Reenactment with Hierarchical Audio-Motion Embedding and Diffusion Interpolation

Haiyang Liu1,  Xingchao Yang2,  Tomoya Akiyama2,  Yuantian Huang2,  Qiaoge Li,  Shigeru Kuriyama2,  Takafumi Taketomi2 
1The University of Tokyo, 2CyberAgent AI Lab
🤗 Huggingface Space (Try it!) arXiv Video Data Code

Video 0

Video 1

Video 2

Video 3

Video 4

Video 5

Video 6

Video 7

Video 8

Video 9

Show-Oliver
60s Reference Video + Unseen Target Speech
Reference Image
Co-Speech Gesture Video Generated from TANGO