TANGO generates co-speech gesture videos by mapping audio features to body motion using hierarchical audio-motion embedding and diffusion interpolation. The model takes speech input and produces corresponding gesture animations, enabling video reenactment with realistic body language synchronized to audio.