Articulate AnyMesh enables creation of movable 3D objects from static meshes using natural language instructions, eliminating category restrictions of traditional methods.

• Automatically identifies parts, joints, and motion constraints from rigid 3D meshes
• Leverages LLMs to understand natural language requests for articulation
• Achieves 86% success rate in creating functional articulated objects
• Operates without category restrictions, handling diverse objects from furniture to toys

This breakthrough enables on-demand creation of interactive 3D assets for engineering applications, virtual environments, simulations, and physical robot interaction testing - all without specialized training data.

Articulate AnyMesh: Open-Vocabulary 3D Articulated Objects Modeling