UDE: A Unified Driving Engine for Human Motion Generation

starting pose end pose a LA Hiphop music 'man walks along then bends down and picks something up.' UDE Figure 1. Our shared Unified Driving Engine (UDE) can support both text-driven and audio-driven human motion generation. Left shows an example of a motion sequence driven by a text description while Right shows an example driven by a LA Hiphop music clip.