Our paper titled “SemGes: Semantics-aware Co-Speech Gesture Generation using Semantic Coherence and Relevance Learning” is accepted to be presented at the international conference on computer vision (ICCV) 2025! This work introduces a novel approach for generating co-speech gestures that are semantically coherent and relevant to the spoken content, improving the naturalness and expressiveness of multimodal communication.

The paper demo is available here.