Stability AI and Tripo AI, two innovative companies in the field of artificial intelligence and 3D modeling, have collaborated to develop TripoSR – a cutting-edge open-source AI model that transforms single source images into high-quality 3D models in less than a second. This groundbreaking technology is specifically tailored to meet the increasing needs of professionals in the entertainment, gaming, industrial design, and architecture industries.
TripoSR utilizes advanced algorithms to generate a textured 3D mesh from a single source image, automatically creating both geometry and textures that are not visible in the original image. Trained on a subset of the Objaverse research dataset containing 3D models under a Creative Commons CC BY license, TripoSR sets itself apart from other image-to-3D reconstruction AI models like OpenLRM and One-2-3-45. In comparative tests conducted by Stability AI and Tripo AI, TripoSR demonstrated superior accuracy in generating models while maintaining efficient processing times.
For 3D artists seeking practical applications of this technology, demonstrations by Tripo AI showcase impressive results with complex models, while Stability AI’s blog post offers insight into more typical outputs. Initial user feedback suggests that TripoSR’s capabilities are particularly promising for creating background models for games or AR applications with hard surface objects. Although performance with organic characters is still being explored.
TripoSR is made available on GitHub under an MIT license along with its dependencies list, while model weights can be accessed on Hugging Face. Despite its advanced capabilities, the model can operate on machines without GPUs with default settings requiring approximately 6GB of VRAM per single image input.
