RoboGesture: Real-Time Co-Speech Gestures on Unitree G1

Loading video
Loading videoECCV 2026 system streams audio into a diffusion-based motion generator to give Unitree G1 real-time co-speech gestures, with an MPC safety filter adding just 5.6 ms per frame.
Category: humanoid
Author: @techniahqrobot
Date: 2026-09-03T00:00:00
Duration: 12.515s
Reference: https://arxiv.org/abs/2608.28693





