MIT, Google and Northeastern unveil InstructMesh for editing AI‑generated 3D models
InstructMesh is a new interface that lets users generate, edit, and 3D‑print objects with AI assistance. Developed by MIT’s CSAIL, Google, Microsoft’s TRELLIS system and Northeastern University, the tool pairs the text‑and‑image‑aware TRELLIS generator with the reasoning power of GPT‑4. Users can prompt a design—such as glasses or a mug—then highlight parts to refine, with sliders for precise…
Key points
- InstructMesh combines Microsoft TRELLIS 3D generator with GPT‑4 to enable natural‑language editing of AI‑created models
- Nearly 80% of generated models were structurally flawed, but novices fixed them about 90% of the time
- Researchers demonstrated functional items such as a dragon‑handle mug, a shell‑shaped whistle, and a shrimp‑like bristle bot
In tests, the system recreated popular Thingiverse models, finding that nearly 80 percent of the AI‑generated shapes were structurally flawed. Novice participants were able to spot and fix those issues about 90 percent of the time, producing functional items like a dragon‑handle mug, a shell‑shaped whistle, and a “bristle bot” shrimp‑like robot. Researchers see the approach extending to AR‑guided on‑the‑fly printing and physics‑based simulations, though those capabilities are still speculative.
The work was presented at the ACM Symposium on User Interface Software and Technology and was supported in part by Google and the MIT‑HPI Collaborative Research Program.
New tool lets users repair AI-generated 3D models, then fabricate them just the way they want
MIT News on AI · 1 October 2026
Loading the full article…
This text was published by MIT News on AI and written by Alex Shipps | MIT CSAIL. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- AWS Bedrock AgentCore Runtime Instances enables multi-agent music production pipeline · 2 src
- Meta's Muse AI agent app reaches #1 on App Store and compiles lists of vulnerable groups on request · 8 src
- Amazon releases Strands Decider 2B, a Jev-inspired open-source decision model · 2 src
- Decagon launches Voice 3 agent with Chord speech model for multilingual calls · 2 src
- Nvidia researchers improve AI agent reliability with a judging model · 1 src
Comments
via GitHub Discussions