Home
AutoN unveils ABot-World Studio, a general world model workshop enabling real-time interaction for AI digital worlds
Autonavi has officially unveiled the General World Model Workshop, known as ABot-World Studio. This platform’s standout feature is the seamless fusion of interactive video generation with 3D scene construction. By simply providing text or image prompts, users can rapidly create highly realistic, AI-driven digital environments that support real-time interaction.
Technically, the studio represents a significant architectural leap. It supports local deployment on a single 5090 GPU and overcomes previous constraints regarding inference duration. Official tests demonstrate stable continuous runtime exceeding one hour, effectively resolving the limitation of earlier models that could only produce short clips.

ABot-World Studio combines two core models, ABot-World0 and ABot-3DWorld0, to address long-sequence video degradation while natively outputting 3DGS spatial assets with accurate geometric structures. This transforms generated content from mere pixel streams into "physical entities" governed by spatial depth and physical rules, enabling users to freely explore photo-quality scenes as if navigating a video game.
To enhance exploration, the platform introduces a distinctive "Time-Space Portal" mechanism. Users can seamlessly transition between distinct 3D scenes, linking previously isolated environments into an expansive network via "space teleportation." For instance, a wooden door in a Jiangnan water town can directly connect to a cyberpunk city.
The ABot-World series models are now fully open source and available on platforms like Reactor and Hugging Face. This innovation holds significant potential for embodied intelligence training, film storyboarding, cultural tourism, and education, empowering users to evolve from passive "content viewers" into active "participants" within digital worlds.
Related article
Xiaomi unveils new AI strategy to build future intelligent infrastructure and reshape the industry
As smart technology advances at an unprecedented pace, Xiaomi is recalibrating its strategic roadmap to harness the power of large AI models. The company recently restructured the engineering team behind Xiaoi, aiming to embed AI deeply into every fa
Besxar Builds Orbital Semiconductor Factory With SpaceX Rockets
Manufacturing in space and returning products to Earth currently relies on a narrow set of approaches: waiting for access to the International Space Station or partnering with select startups that launch spacecraft capable of returning payloads from
Liang Wenfeng on Open Source: 6x Profits Without Overtime, 100x Gains Tied to Open Source
Following a recent leak of internal discussions from DeepSeek’s investor exchange, founder Liang Wenfeng clarified the strategic reasoning behind their open-source approach. He firmly stated, “We will definitely open-source, including our most powerf
Related Special Topic Recommendations
Comments (0)
0/500
Autonavi has officially unveiled the General World Model Workshop, known as ABot-World Studio. This platform’s standout feature is the seamless fusion of interactive video generation with 3D scene construction. By simply providing text or image prompts, users can rapidly create highly realistic, AI-driven digital environments that support real-time interaction.
Technically, the studio represents a significant architectural leap. It supports local deployment on a single 5090 GPU and overcomes previous constraints regarding inference duration. Official tests demonstrate stable continuous runtime exceeding one hour, effectively resolving the limitation of earlier models that could only produce short clips.

ABot-World Studio combines two core models, ABot-World0 and ABot-3DWorld0, to address long-sequence video degradation while natively outputting 3DGS spatial assets with accurate geometric structures. This transforms generated content from mere pixel streams into "physical entities" governed by spatial depth and physical rules, enabling users to freely explore photo-quality scenes as if navigating a video game.
To enhance exploration, the platform introduces a distinctive "Time-Space Portal" mechanism. Users can seamlessly transition between distinct 3D scenes, linking previously isolated environments into an expansive network via "space teleportation." For instance, a wooden door in a Jiangnan water town can directly connect to a cyberpunk city.
The ABot-World series models are now fully open source and available on platforms like Reactor and Hugging Face. This innovation holds significant potential for embodied intelligence training, film storyboarding, cultural tourism, and education, empowering users to evolve from passive "content viewers" into active "participants" within digital worlds.
Xiaomi unveils new AI strategy to build future intelligent infrastructure and reshape the industry
As smart technology advances at an unprecedented pace, Xiaomi is recalibrating its strategic roadmap to harness the power of large AI models. The company recently restructured the engineering team behind Xiaoi, aiming to embed AI deeply into every fa
Besxar Builds Orbital Semiconductor Factory With SpaceX Rockets
Manufacturing in space and returning products to Earth currently relies on a narrow set of approaches: waiting for access to the International Space Station or partnering with select startups that launch spacecraft capable of returning payloads from
Liang Wenfeng on Open Source: 6x Profits Without Overtime, 100x Gains Tied to Open Source
Following a recent leak of internal discussions from DeepSeek’s investor exchange, founder Liang Wenfeng clarified the strategic reasoning behind their open-source approach. He firmly stated, “We will definitely open-source, including our most powerf











