Stability AI unveils TripoSR, an AI tool that generates 3D models from images

What sets TripoSR apart is its remarkable speed and accessibility. When tested on an Nvidia A100 GPU, it can generate draft-quality 3D outputs, complete with textured meshes, in a mere 0.5 seconds, outperforming other open image-to-3D models like OpenLRM.

But the true power of TripoSR lies in its versatility. Even users without access to powerful GPUs can harness its capabilities, as the model is optimised to run under low inference budgets, according to the press release.

Building upon the recent work of LRM (Large Reconstruction Model For Single Image to 3D), the team at Stability AI and Tripo AI have introduced several technical improvements to the base model. These enhancements include channel number optimisation, mask supervision, and a more efficient crop rendering strategy, all of which contribute to the model’s impressive performance.

In keeping with the spirit of open-source collaboration, the model weights and source code for TripoSR are available for download. The code can be accessed from Tripo AI’s GitHub while model weights are available on Hugging Face.