AIsphere Opens Hong Kong Cyberport Office: PixVerse R2 Debuts, AI Industry Co-Creation Program Launched

AI video large model company AIsphere recently announced the official opening of its Hong Kong office at Cyberport. This is not merely the establishment of an office space, but is positioned as an AI hub radiating to local enterprises, government agencies, universities, and creators — it began operations on the day of the opening and plans to recruit more local employees and continue hosting industry-specific events.

From Office Space to "AI Hub": The Logic Behind the Hong Kong Layout

At the opening ceremony, AIsphere co-founder and President Xie Xuzhang stated that Hong Kong possesses an international perspective, a solid research foundation, and a vibrant creative industry, making it an important stop in the company's global development. He explicitly articulated the role of "an AI hub connecting industries across Hong Kong" — not just bringing technology in, but truly applying AI video and real-time world model capabilities to industry and creation.

The backdrop to this statement is the Hong Kong SAR government's recently announced Policy Address and its first five-year plan (2026–2030), which explicitly proposes empowering all industries with artificial intelligence, continuing to advance the "AI+" initiative, accelerating computing infrastructure development, and enhancing citizens' AI literacy. As one of the first AI video large model companies to formally establish an office in Hong Kong, AIsphere is tapping into a direction with clear policy support.

The lineup of guests at the ribbon-cutting ceremony reflected this: Acting Director-General of Investment Promotion Jimmy Liu, Cyberport CEO Dr. Rocky Cheng, Associate Director-General of the Office for Attracting Strategic Enterprises (OASES) Yang Shuyuan, and representatives from the Hong Kong Productivity Council were all present. In his remarks, Liu stated that AIsphere's establishment "fully demonstrates Hong Kong's attractiveness in the AI field both domestically and internationally," and specifically noted he was pleased to see the company's plans to participate in local talent development and community AI literacy improvement.

Digital-Intelligent Hong Kong · AIsphere AI Industry Co-Creation Program: First Batch of Partnerships Landed

On the opening day, AIsphere announced the launch of the "Digital-Intelligent Hong Kong · AIsphere AI Industry Co-Creation Program" and signed multiple strategic cooperation documents with Hong Kong tech enterprises, higher education institutions, and youth social organizations. These cooperation directions are worth examining in detail.

At the industry level, the company established a strategic partnership with spatial intelligence company Qiyu Innovation, integrating real-world 3D reconstruction with AI video generation technology to explore joint solutions for film and television, cultural tourism, real estate, and other industry scenarios, with plans to jointly expand into overseas markets.

At the education and research level, the company is collaborating with The Hong Kong Polytechnic University to introduce AI video creation into humanities teaching, combining Chinese history and culture courses to explore teaching practices in responsible AI and digital storytelling, and on this basis expanding into talent development, student internships, and joint research.

At the social service level, AIsphere is partnering with the Youth Committee of the Hong Kong Fujian Sanming Association to conduct AI popular science lectures, creative experience workshops, and tech enterprise visits for local residents and youth, and to jointly carry out youth exchange and AI cultural co-creation activities such as "AI Looks at Sanming" through networks in both Fujian and Hong Kong. Meanwhile, the company is also collaborating with local organization Hong Kong Youth New Ideas to explore applications of AI technology in community building, community services, and youth exchange — for example, using AI video to produce anti-fraud public welfare content and supporting youth exchange activities along the "Belt and Road."

AIsphere's Chief Representative in Hong Kong, Pan Yuan, interpreted this as follows: "Systematic popular science displays and interactive experiences can enhance the public's rational understanding of AI, create a more inclusive and stable social environment for technology applications, and solidify the social foundation for AI's long-term development in Hong Kong."

PixVerse R2 Live Demo: What Can a Real-Time World Model Do

At the opening ceremony, AIsphere showcased the recently released general-purpose real-time world model PixVerse R2 to guests and media in the exhibition hall. According to the on-site introduction, this is the world's first general-purpose real-time world model with Scaling capability.

Unlike traditional AI video with its "pre-recorded, fixed-duration, one-way output" format, a general-purpose real-time world model can perform real-time generation and interactive modeling of visuals, audio, and other information, enabling video streams to change synchronously as users input text, voice, images, or operational commands. Its technology relies on native multimodal modeling, autoregressive streaming generation, and an instantaneous response engine, balancing cross-modal understanding, long-sequence consistency, character and scene state continuity, and low-latency feedback to achieve "generating while interacting, evolving while generating."

Guests experienced modifying character states, environmental changes, camera directions, and narrative branches in real time through natural language commands across multiple scenarios including games, interactive film and television, and digital humans. The model can not only "live with memory" but also "understand various languages," responding to every user action with causally appropriate reactions — moving from "watching" to real-time interaction where "words become reality."

Xie Xuzhang told Leiphone that the company will continue to explore deep integration of multiple flagship models including PixVerse R2 into locally advantageous industries. Specifically, in the film and cultural creativity sector, the PixVerse series models' cinematic visuals, multi-shot scheduling, and character consistency capabilities can help creators transform scripts, characters, and worldviews into discussable, verifiable imagery during the pre-development stage; in cultural tourism, the real-time world model can generate interactive immersive guided tours and personalized experience content for scenic areas, museums, and cultural tourism projects; in advertising, marketing, and e-commerce, brand content can instantly generate personalized interactive visuals based on user feedback, and audiences can transform from bystanders into narrative participants in interactive film-games.

What Is a World Model?

A world model is a frontier concept that has emerged in the AI field in recent years, referring to enabling a model to internally construct a computable representation of an external environment (such as the physical world or a virtual scene), thereby understanding, predicting, and even simulating how the environment changes over time and with actions. Unlike traditional image generation or video generation models that focus only on "whether a single frame looks good," world models emphasize modeling the laws of state evolution — that is, "if I operate this way, what will the world look like in the next second." This capability is a shared underlying requirement for scenarios including autonomous driving, robotics, game engines, and interactive film and television. Meta, Google DeepMind, NVIDIA, and other institutions all regard world models as one of the key paths toward artificial general intelligence (AGI). The "general-purpose real-time world model" proposed by AIsphere emphasizes achieving this interactive, predictable, and real-time responsive capability in the video dimension.

What Does "Scaling Capability" Mean?

Scaling capability originates from the famous "Scaling Law" in the large model field: as model parameters, training data, and compute increase, model capabilities show predictable, continuous improvement. A model "having Scaling capability" means its architecture design allows for stably trading more parameters and compute for stronger performance without quickly hitting a performance ceiling. For real-time world models, Scaling is especially difficult — because real-time inference must be completed under low-latency constraints, the model must be both "large" and "fast." AIsphere's claim that PixVerse R2 is the world's first general-purpose real-time world model with Scaling capability is essentially asserting that it has found an architectural solution that balances real-time performance and scalability, which is somewhat pioneering on the technical path.

AIsphere's Hand: User Scale and Funding Pace

AIsphere was founded in 2023, focusing on the R&D and application of multimodal video large models and real-time world models. The company's self-developed PixVerse series video large models include the V series general-purpose video models, C series vertical film and television models, and R series general-purpose real-time world models, with overall capabilities positioned in the industry's first tier.

According to disclosures, the company's PixVerse and its mainland China market "PaiWo AI" app and open platform currently cover 177 countries and regions worldwide, with cumulative registered users exceeding 170 million. In July of this year, the company completed its Series C funding round, with cumulative funding of nearly 3 billion RMB in this round. Investors include CDH Investments, Alibaba, Lollapalooza Capital (Wang Huiwen's family office), Ivy Capital, Huiyuan Capital, Zhongding Capital, Korea Investment Partners, Lion X Fund under the OCBC ecosystem, BlueFocus, CloudAlpha, iGlobe Partners, and more than ten other domestic and international institutions.

Previously, AIsphere had explored directions including short dramas, animation, variety show visuals, and interactive film-games with content partners such as Mango Media in mainland China, and established partnerships with industry partners including Alibaba, Giant Group, China Ruyi, BlueFocus, and 37 Interactive Entertainment. The cross-regional transferability of these experiences will be a key variable in determining whether its Hong Kong layout can produce substantive output.

Why This Move Is Worth Watching

Hong Kong's positioning in the AI field in recent years is evolving from "research base" to "industry implementation hub." Cyberport CEO Dr. Rocky Cheng's statement is quite representative: "Cyberport is committed to building a vibrant AI and digital technology ecosystem. We welcome AIsphere to Cyberport, and believe its technical accumulation in video generation and real-time world models will bring new inspiration and collaboration opportunities to our community."

AIsphere's choice to land in Hong Kong at this time hits the policy window while leveraging Hong Kong's international attributes to pave the way for going global — the Qiyu Innovation cooperation explicitly mentions "jointly expanding into overseas markets," and PixVerse already covers 177 countries and regions. The real-time world model category is still in its early stages, and whoever can first establish a commercial closed loop in film and television, cultural tourism, advertising, and other scenarios will be more likely to define the rules of this track. The opening of the Hong Kong office is a concrete action by AIsphere to bet on this window of opportunity.

The Role of Hong Kong Cyberport and the "Office for Attracting Strategic Enterprises"

Cyberport is a digital technology flagship campus wholly owned by the Hong Kong SAR government. Since its launch in 2004, it has been an incubation base for local startups and digital technology companies. In recent years, it has explicitly designated artificial intelligence, Web3, and smart living as its three strategic directions, providing tenants with rent subsidies, computing resources, and industry matchmaking services. The Hong Kong Office for Attracting Strategic Enterprises (OASES) is a dedicated agency established after the 2022 Policy Address, responsible for proactively attracting high-potential global tech enterprises to settle in Hong Kong and providing one-stop support including settlement packages, financing matchmaking, and talent introduction. AIsphere simultaneously obtaining physical space from Cyberport and official endorsement from OASES means its landing is not just a market action but also carries clear policy support attributes — for an AI large model company that needs continuous investment in compute and talent, this is a key condition for reducing early-stage costs and accelerating localization.

One-line Summary

AIsphere opens its Hong Kong office at Cyberport, launches the AI Industry Co-Creation Program, and showcases the real-time world model PixVerse R2.

Paragraph Summary

AIsphere has officially opened an office at Hong Kong Cyberport, positioned as an AI hub radiating to local enterprises, government, universities, and creators, beginning operations on the opening day. The company launched the "Digital-Intelligent Hong Kong · AIsphere AI Industry Co-Creation Program," signing multiple partnerships with The Hong Kong Polytechnic University, Qiyu Innovation, and the Youth Committee of the Hong Kong Fujian Sanming Association, covering industry, education and research, and social services. The general-purpose real-time world model PixVerse R2 showcased on-site supports real-time modification of character states, environmental changes, and narrative branches through natural language commands. The PixVerse series currently covers 177 countries and regions with over 170 million cumulative registered users, and completed nearly 3 billion RMB in Series C funding in July this year. This layout aligns with Hong Kong's "AI+" policy direction and provides an international springboard for overseas expansion.