Visko Launches Orbis Real-Time Video Model With $10M Pre-Seed Round

Sunnyvale-based Visko Platform Inc. has launched Orbis, a foundation model capable of streaming 4K video in real time. The release coincides with a $10 million pre-seed funding round led by Llama Ventures.

Orbis distinguishes itself from traditional AI video generators by operating as a "live model." While standard systems render static clips after a prompt, Orbis generates video continuously at 24 frames per second in 4K resolution. The model maintains persistent memory and interactivity, allowing users to update prompts mid-run. Technical benchmarks indicate that Orbis leads other real-time and long-video systems in aesthetic and technical quality on the DOVER scale, as well as visual and motion quality on VideoAlign. Furthermore, the system can sustain hour-scale generation without experiencing the quality or color drift common in long-form AI video.

A Shift in Video Generation

Traditional AI video generation typically functions as a request-and-wait process, where a user provides a prompt and the model renders a pre-defined sequence. Visko’s approach treats video generation as a continuous process, mirroring the next-token prediction used in large language models (LLMs). This creates a "world model" that runs in step with the environment rather than producing a fixed file.

Founded in 2025 by former Apple researcher and Stanford Ph.D. Qing (Will) Yin, Visko is built by a 16-person team featuring alumni from Google DeepMind, Meta, Amazon, Tesla, and Apple. Yin describes this shift as a fundamental change in the technology, stating, "The next paradigm of AI is not a model you call, but a process that runs in step with the world itself."

Implications for Robotics and Simulation

The ability to generate steerable, physically grounded 4K video in real time has immediate applications for autonomous systems. By creating synthetic environments, developers can simulate "corner cases"—rare or dangerous scenarios—that would be too expensive or hazardous to stage in the physical world. This capability is intended to accelerate the training of general-purpose robotics and the testing of autonomous vehicles.

Michael I. Jordan, chairman of Visko's advisory board, noted that the company is pursuing "live, stateful models that continuously perceive, predict, and interact with the world in real time while maintaining physical consistency."

The Path Forward

As Visko deploys its pre-seed capital, the industry will be watching how Orbis scales beyond controlled simulations into broader industrial applications. While the technical benchmarks show leadership in motion and aesthetic quality, the primary challenge remains the integration of these live models into real-world robotic hardware to see if the synthetic training translates to physical reliability.