LAVA Lab and VCL Lab presented a live demo titled “Situately: Embodied Conversational Agent with LLM-Powered Nonverbal Behavior and Object Interaction” at the IEEE VR conference (March 21–25, 2026) in Daegu.
Situately is a real-time embodied conversational agent framework that enables coordinated verbal and nonverbal interaction within a shared virtual environment. Leveraging large language models and ongoing conversational context, the system generates structured outputs that control avatar-level behaviors, including gestures, gaze, and object interaction.
These outputs are directly mapped to real-time animations, allowing the agent to exhibit socially and spatially grounded behavior beyond verbal communication. The demo showcased how concurrent multimodal control enhances users’ sense of social presence and the perception of being “together” with the agent in the same space.
