InteractAnything: Zero-shot Human Object Interaction Synthesis via LLM Feedback and Object Affordance Parsing

Diverse human interactions. (b) Detailed interactions synthesis of open-set objects. (c) Novel interactions given generative objects. A person holds/pulls/lies on the chair… A person grasps the backpack/table/ball/knife… A person lifts/rides/craddles the tesla car/motorcycle/horse/baby… Figure 1. 3D human object interaction synthesis by InteractAnything. Given a simple text description with goal interaction and any object mesh as input, our method enables the generation of diverse, natural, detailed, and novel interactions for open-set 3D objects in a zero-shot manner. The orange and green boxes of (b) indicate detailed contact poses from different views.