AgentHOI: Multi-Agent Reasoning for Human-Object-Interaction Video Generation via Implicit Representation Alignment

Researchers introduced AgentHOI, a text-driven human-object interaction (HOI) video generation method.

It uses multi-agent reasoning to plan perception, interaction, and motion before generating video.

The system also learns implicit text-motion alignment, reducing reliance on explicit motion inputs at inference.

This makes HOI video synthesis more controllable, scalable, and practical for complex interactions.