TripoSR is an open-source 3D reconstruction model from VAST AI Research (developed with Stability AI) that turns a single 2D image into a textured 3D object in under a second. It has around 7,200 stars on GitHub and was updated in 2026.
Built on the Large Reconstruction Model (LRM) architecture, TripoSR takes one photograph or illustration and predicts a complete 3D mesh — geometry plus appearance — in a single feed-forward pass, with no per-object optimization. That speed changes what is practical: interactive 3D content creation, rapid prototyping for games and AR, and on-demand asset generation from product photos all become feasible. The model was trained with careful data curation and rendering improvements that give it an edge in reconstruction quality over earlier single-image methods, and the released code includes inference scripts and examples so you can generate meshes from your own images right away. TripoSR is the open research release behind the Tripo product line and is published under the MIT license.
TripoSR is completely free and open source under the MIT license, including the model weights. Commercial 3D generation services exist separately, but everything in this repository can be used at no cost.
TripoSR is for game developers, 3D artists, AR/VR builders and researchers who need fast 3D assets from images, and for anyone experimenting with single-image 3D reconstruction without a commercial API.