New model synthesizes human-object interaction with real-time control

Researchers have developed a novel real-time human-centric world model designed for upper-body human-object interaction synthesis. This model combines continuous control for human motion, including body, hands, and face, with discrete control for object interactions, using language-encoded states like 'grasp' or 'no contact'. The system achieves efficient streaming inference at 25 FPS on two H100 GPUs, demonstrating improved motion fidelity and realistic hand-object coordination. AI

IMPACT Enables more realistic and controllable synthesis of human-object interactions in real-time applications.

RANK_REASON Research paper detailing a new model for human-object interaction. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• H100 GPUs

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →