TripoSR
An open-source image-to-3D model that runs locally with Python and PyTorch. MIT-licensed code and weights, with about 6GB VRAM for default inference.
TripoSR reconstructs a 3D object from a single image on your own hardware. Developed by Tripo AI and Stability AI, it's an open-source model for researchers, developers, and artists who want image-based 3D generation they can use in their own projects. The MIT license covers both the source code and pretrained models.
Its focus is fast reconstruction. TripoSR builds on the Large Reconstruction Model (LRM) approach and generates a model in a single forward pass. The reported generation time is under half a second on an NVIDIA A100 GPU, so that speed reflects a specific hardware setup. Local inference uses Python and PyTorch, with CUDA acceleration where available; the default settings use about 6GB of VRAM for one input image.
The output can use vertex colors or a baked texture, and TripoSR can process multiple input images in a batch. It saves reconstructed models locally, giving developers a way to incorporate image-to-3D generation into their own software. An interactive online demo also provides a way to try the model.
TripoSR is distinct from Tripo's hosted browser service. That service accepts text prompts as well as images and includes quad mesh generation, PBR texturing, and automatic rigging. Those are capabilities of the online platform; the local TripoSR codebase focuses on reconstructing objects from single images.
