Alibaba Group's AutoNavi has officially launched the general world model workshop ABot-World Studio, and simultaneously opened it for public testing. This platform first integrates interactive video generation and 3DGS scene generation into a single product. Users need only input a text or upload an image to generate an AI world that can be interacted with in real time and freely shared, supporting the export of videos and 3DGS files.
ABot-World Studio has made breakthroughs in deployment and inference capabilities, enabling local deployment on a single NVIDIA RTX5090 graphics card, with no limit on inference duration set in the architecture design. Official tests show that a single continuous inference can run stably for more than one hour without crashing or quality degradation, significantly breaking through the current industry's one-minute generation time limit. In addition, the product innovatively includes a "Time and Space Portal" function, allowing users to freely move between different complete 3D worlds, further enhancing the immersive interaction experience.
This capability is based on the underlying ABot-World series of general interactive world models, including ABot-World0 for video generation and ABot-3DWorld0 for 3D generation. Both models can perform local inference on a single consumer-level GPU, without relying on dedicated computing power, which is rare in the current world model field. Currently, all related models are fully open-sourced.
In terms of application, ABot-World Studio can provide simulation training environments for embodied intelligent robots, support content creation scenarios such as games and films, reducing the creative validation cycle from weeks to hours, and can also be applied to fields such as cultural tourism and education, helping users experience history, culture, and virtual spaces from a first-person perspective, further expanding the industrial application space of world models.
