Researchers have developed SICAGE, a novel framework for generating co-speech gestures that are sensitive to cultural nuances and independent of individual speaker styles. This system learns cultural representations from audio and text, ensuring that the generated gestures are culturally appropriate and synchronized with speech. The framework was validated using the newly created TED4C-L dataset, which comprises 106 hours of gesture data from 764 TED speakers across four cultural groups. Experiments demonstrated that SICAGE significantly enhances the realism, diversity, synchronization, and cultural consistency of generated gestures. AI
IMPACT This research could lead to more culturally sensitive and realistic AI avatars and virtual assistants.
RANK_REASON The cluster contains a research paper detailing a new AI model and dataset. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
