Runway Develops Technology to Stream AI-Generated Video in Real Time

Runway, a company working on video generation AI, has revealed that it is pursuing a new approach to stream AI-generated video in real time. Whereas traditional AI video generation required users to wait for completion after issuing instructions, Runway's vision is to eliminate that waiting time and deliver a continuous flow of video as users interact with it.

AI-driven video generation has rapidly improved in accuracy over the past few years, making tools that can create video content of several seconds to tens of seconds from text or images widely available. However, in the currently mainstream approach, users cannot confirm results until generation is complete. As videos become longer, waiting time increases, and a common industry-wide challenge is that the pace of trial and error is hindered. Runway's concept is positioned as directly addressing this challenge.

The technical foundation for this approach is the world model called "GWM-1" that Runway has already developed. A world model refers to a mechanism that allows AI to internally learn the rules of physical environments and movements in reality, enabling it to generate video continuously on a frame-by-frame basis. By performing this frame-by-frame sequential generation, GWM-1 has high affinity with streaming-style output, according to Runway's explanation.

Runway views the application of this technology as extending beyond creator-focused video production tools. The company is also considering applications in robotics and autonomous driving, where AI's ability to "understand and generate environments as video" in real time could be combined with sensors and control systems. This suggests a direction toward expanding video generation from a "tool for expression" to "infrastructure for perception and decision-making."

The significance of this development for the industry cannot be limited to improved usability alone. If users can intervene in video generation in real time, the workflow of video production itself could be fundamentally transformed. Furthermore, if applications in robotics and autonomous driving are realized, the market for video generation AI could become a turning point where it expands from the scope of "content creation tools" to broader industrial applications.

At this point, Runway has not announced specific timelines or detailed specifications for service deployment, and the stage is one of watching the progress of technological development and practical implementation. How GWM-1 will be further developed and how the new experience of streaming generation will be delivered to users remains to be seen, and continued monitoring of developments is warranted.