7 datasets found

TED Expressive is a recent English-based dataset for co-speech gesture generation. Different from TED Gesture dataset that only contains body skeletons, TED Expressive extracts...

TED Gesture is a large-scale English-based dataset for co-speech gesture generation, composed of 1766 TED videos from various topics and different narrators.

Co-speech gesture generation is crucial for automatic digital avatar animation. However, existing methods suffer from issues such as unstable training and temporal...

The GENEA Challenge 2022 dataset is a large-scale open challenge for data-driven automatic co-speech gesture generation. It consists of 18 hours of full-body motion capture,...

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

A dataset for co-speech gesture generation, containing 35,758 clips.

A large-scale co-speech gesture dataset containing more than 40M high-quality 3D meshed postures across 4.3K speakers from in-the-wild talk show videos.