Project Continuum is an experimental application from Runway's Runway Labs that explores real-time video in an operating-system-style interface. Runway presented an early demonstration on October 1, 2026.[1]

Runway Labs is the company's internal incubator for experimental applications of generative video and world models. Runway announced the unit on March 11, 2026, with co-founder and Chief Innovation Officer Alejandro Matamala Ortiz as its leader. Its remit includes prototyping uses of these models beyond established media-production workflows.[2]

Demonstrated interactions

Area	Demonstration
Portals	Selecting onscreen content opens an interactive visualization. Resizing it prompts scene regeneration; blurred edges distinguish generated material.[3]
Visual Thinking	An agent uses video to visualize a requested 3D-printable bicycle mount.[3]
Responsive Video Interfaces	Suggestions reflect the material currently open in the application.[3]
Interactive Worlds	A suggested Mars outpost opens as a full-screen scene that responds to clicks.[3]

The prototype also has a prompt bar, tabs for previous tasks, and a drawer for reopening generated portals.[3]

Relationship to Solaris

Runway places Continuum alongside its Solaris interface world model.[1] Announced on August 31, 2026, Solaris generates interface frames in response to user input. Runway describes a division of work in which a language model interprets requests and determines scene behavior, while Solaris renders the interaction.[4]

Solaris builds on Gen-4.5. Its adaptation uses autoregressive frame generation, reduces the number of denoising steps, and trains the fast model on its own outputs. These are the company's descriptions of the rendering model, rather than a complete technical specification of the Continuum application.[4]

Runway identifies unresolved Solaris problems with legible text, factual grounding, consistency during long sessions, and accessibility integration. These qualifications matter for interfaces used to teach, convey product information, or support assistive technologies.[4]

Related interface research

Generative interfaces also appear in academic work that produces conventional interface code. Jiaqi Chen and colleagues' Generative Interfaces for Language Models, published in the Findings of ACL 2026, studies task-specific interfaces as an alternative to conversational responses. Its evaluation considers functional, interactive, and emotional aspects of user experience.[5]

The researchers' method maps a query into an intermediate representation, decodes that representation into interface code, and refines the result iteratively.[6] This differs from the frame-generation approach described for Solaris.[4] The paper concerns a separate system and does not evaluate Continuum.[5]

Preview and availability

The October announcement presents an early research demonstration. It does not specify a Continuum download, price, or public release date.[1][3] Runway separately offers an early-access request form for Solaris.[4]

References

Editing: Project Continuum