1 dataset found

Groups: Co-speech gesture generation

Filter Results
  • TalkSHOW

    A high-quality audiovisual dataset for speech-to-motion generation, consisting of 3D holistic body meshes and synchronized speech.