InstructMesh AI tool enables functional repairs for 3D-printed designs

Researchers from MIT, Google, and Northeastern University have developed InstructMesh, an AI-driven tool designed to bridge the gap between visually realistic 3D models and functional, printable objects. While generative AI can create impressive 3D designs from text or images, these models often lack geometric correctness or structural integrity, such as a mug with a sealed opening or walls that are too thin to support weight.

InstructMesh allows users without professional CAD experience to identify and fix these flaws using natural language instructions or simple parametric sliders. The system integrates Microsoft’s TRELLIS, which generates the initial 3D models, with the GPT-4 large language model to interpret and execute user edits.

In testing, the tool successfully produced functional items, including a dragon-themed mug, a shell-shaped whistle, and a customized knee brace. The technology also shows potential for engineering and medical applications, such as creating specialized robotic shells or personalized medical devices.