Grounded Gesture Generation: Language, Motion, and Space

TL;DR

This paper introduces a multimodal dataset and framework for grounded gesture generation that integrates speech, motion, and spatial context, aiming to improve embodied, communicative agents.

Contribution

It provides a new synthetic and VR-based dataset with synchronized multimodal data and a framework connecting gesture modeling to physics simulation for grounded interaction.

Findings

Over 7.7 hours of multimodal data collected

Standardized in HumanML3D format for consistency

Framework enables situated evaluation of gestures

Abstract

Human motion generation has advanced rapidly in recent years, yet the critical problem of creating spatially grounded, context-aware gestures has been largely overlooked. Existing models typically specialize either in descriptive motion generation, such as locomotion and object interaction, or in isolated co-speech gesture synthesis aligned with utterance semantics. However, both lines of work often treat motion and environmental grounding separately, limiting advances toward embodied, communicative agents. To address this gap, our work introduces a multimodal dataset and framework for grounded gesture generation, combining two key resources: (1) a synthetic dataset of spatially grounded referential gestures, and (2) MM-Conv, a VR-based dataset capturing two-party dialogues. Together, they provide over 7.7 hours of synchronized motion, speech, and 3D scene information, standardized in the HumanML3D format. Our framework further connects to a physics-based simulator, enabling synthetic data generation and situated evaluation. By bridging gesture modeling and spatial grounding, our contribution establishes a foundation for advancing research in situated gesture generation and grounded multimodal interaction. Project page: https://groundedgestures.github.io/