Researchers have developed GLARE, a new system for generating realistic listener reactions in dyadic conversations. Existing methods struggle with natural listener behavior, lacking appropriate reaction annotations and evaluation metrics. GLARE addresses this by introducing a new dataset with over 147 hours of paired speaker-listener videos and 64,557 reaction annotations across six categories. The system utilizes a flow-matching transformer conditioned on Qwen2-Audio prosody and a temporal reaction loss. A novel evaluation protocol measures reaction occurrence, temporal alignment, and visual quality, demonstrating GLARE's superiority over prior methods in both visual fidelity and behavioral realism. AI
IMPACT This research could lead to more natural and engaging AI-powered conversational agents and virtual assistants.
RANK_REASON The cluster contains a research paper detailing a new system and dataset for generating realistic listener reactions in conversations. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- Gotit.pub
- Hugging Face
- Qwen2-Audio
- RealTalk
- ScienceCast
- Seamless interaction in space
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →