TripoSR

TripoSR is an open-source single-image-to-3D reconstruction model jointly developed by Tripo AI and Stability AI. Feed it one image and it produces a 3D mesh, and it is designed to run locally on a GPU. It fits 3D developers and researchers comfortable with their own machines who want to spin up meshes quickly for prototyping.

Its biggest strengths are speed and cost. Inference finishes in under a second on a GPU, pushing the generation-speed metric to 92, and the MIT license makes it free even for commercial use, lifting the value score to 90. Being able to download the code and run it unlimited in your own environment with no usage fees is the core appeal of an open-source model. Output is delivered as an OBJ mesh.

The limits lie in scope. There is no rigging or texturing, so adding a skeleton or surface detail to the generated mesh requires post-processing in other tools. There is no Korean support either.

In short, for developers who handle a GPU directly and need fast, cost-free mesh generation, it is a strong pick. If you want rigged, textured, finished assets, or an easy install-free tool, it is better to look at other 3D generation services that include post-processing features.