Runway Unveils "Gen-1," a Pioneer in AI-Powered Video Style Transfer
Runway, a startup that had been involved in the development of Stable Diffusion, announced its video generation AI system “Gen-1.” Its defining feature is the ability to apply a style or composition specified through text or an image while keeping the structure of the original footage intact. Runway said it would begin with a small group of invited users and open access to the entire waitlist within a few weeks.
Details
- Capabilities: It can transform footage of a person walking through a city into claymation-style puppets, or turn a stack of books on a desk into a nighttime cityscape — changing only the “look” of a video while keeping its underlying structure
- Availability: On launch day, access started with a small group of invited users, with plans to expand to the full waitlist within a few weeks
- Technical positioning: Rather than generating video from scratch, the system specializes in “video-to-video” transformation applied to existing footage
- Evaluation: In internal user studies, the system was preferred 73.5% of the time over Stable Diffusion 1.5 for image-to-image conversion, and 88.2% of the time over Text2Live for video-to-video conversion
What happened next
Gen-1 was the opening move in the video generation AI market — just a month and a half later, in March 2023, “Gen-2” arrived, capable of generating video entirely from text or images alone. Building on Gen-1, Runway established itself as a pioneer in the video generation AI space.