A few days ago,Tencentorigin of the universe 3D The newest member of the World Model Series -- Hunyuan World-Voyager (Hybrid Voyager) was officially released, claiming to be "the industry's first ultra-long roaming world model to support native 3D reconstruction".
The model focuses on the expansion of AI applications in the field of spatial intelligence, and will provide high-fidelity 3D scene roaming capabilities for virtual reality, physics simulation, game development and other fields.
According to the official introduction, Hybrid Voyager breaks through the limitations of traditional video generation in terms of spatial consistency and exploration range, and is capable of generating long-distance, world-consistent roaming scenes, supporting the direct export of video to 3D format.
The "3D Input - 3D Output" feature of Hybrid Voyager is highly compatible with the previously open-sourced Hybrid World Model 1.0, which can further extend the roaming range of the 1.0 model and allow stylized control and editing of the generated scene. Voyager also supports video scene reconstruction, 3D object texture generation, video style customization, video depth estimation, and many other 3D understanding and generation applications.
In terms of technology, for the first time, Hybrid Voyager supports native 3D memory and scene reconstruction by combining space and features; adding 3D conditions on the input side to ensure an accurate view of the screen, and generating a 3D point cloud directly on the output side. Meanwhile, Voyager also introduces a scalable world caching mechanism, based on the initial 3D point cloud cache generated by the 1.0 model, which is projected to the target camera view to provide guidance for diffusion modeling.
It is worth mentioning that the model ranked first in terms of overall capability on the WorldScore, a benchmark test of world models released by Fei-Fei Li's team at Stanford UniversityVoyager outperforms existing open source methods in both video generation and 3D reconstruction tasks. Voyager also achieves better results in both video generation and video 3D reconstruction.
HunyuanWorld-Voyager is now live, the technical report is publicly available, and the source code is freely available on GitHub and Hugging Face.
Project homepage: https://3d-models.hunyuan.tencent.com/world/
Github: https://github.com/Tencent-Hunyuan/HunyuanWorld-Voyager
Hugging Face: https://huggingface.co/tencent/HunyuanWorld-Voyager
Technical report: https://3d-models.hunyuan.tencent.com/voyager/voyager_en/assets/HYWorld_Voyager.pdf