Runway Gen-2
By Runway
Runway Gen-2 is a text-to-video and image-to-video generation model from Runway, offered through Runway's hosted creative platform, capable of producing short video clips from a text prompt, an input image, or a combination of both. It…
Definition
Runway Gen-2 is a text-to-video and image-to-video generation model from Runway, offered through Runway's hosted creative platform, capable of producing short video clips from a text prompt, an input image, or a combination of both. It followed Runway's earlier Gen-1 video-to-video model and was later succeeded by Runway's Gen-3 model, and it is offered as a closed, subscription-based service rather than as downloadable weights.
Overview
Runway Gen-2 is a text-to-video and image-to-video generation model from Runway that addresses the need for a general-purpose, product-accessible way to create short video clips without traditional filming or animation, offered through Runway's hosted creative platform rather than as a research demonstration or open-weight release. The model accepts a text prompt, an input image, or a combination of both as conditioning, and generates a short video clip consistent with that input, giving users more than one entry point into video generation: starting purely from a written description, animating a specific still image, or using an image alongside a text prompt to guide how that image should be extended into motion. This flexibility means a user working from a specific existing photograph or illustration can animate that exact asset rather than needing to redescribe its visual content in words and hope a purely text-conditioned model reproduces it closely, which is a common source of mismatch in text-only video generation workflows. Within the text-to-video landscape, Gen-2's multiple input modes distinguish it from models built around a single conditioning type, such as purely text-conditioned video generators, and its status as a commercial, hosted product accessible through a standard web application differentiates it from open-weight research models like CogVideoX or Make-A-Video, which require more technical setup to run. It is part of Runway's broader suite of AI video tools rather than a single standalone model. In practice, Gen-2 has been used by video creators, marketers, and filmmakers for generating short b-roll-style clips, visual effects elements, and concept previews, often as one step within a larger editing workflow built in Runway's platform, which also includes other generation and editing tools that can be combined with Gen-2's output. Its limitations include the short clip lengths and occasional visual inconsistency typical of the text-to-video generation stage generally, along with being a closed, hosted service whose usage is tied to Runway's platform and pricing rather than being self-hostable or fine-tunable by users. Teams needing longer-form video, offline generation, or full control over the underlying model typically look to open-weight alternatives or traditional production methods for parts of a project Gen-2 cannot adequately cover. Runway has continued to iterate on its video generation models beyond Gen-2, and the platform's broader positioning as a creative tool for professional and semi-professional video work, rather than as a research artifact, has shaped its feature set toward practical production concerns such as consistent output formats and integration with standard editing workflows.
Key Concepts
- Generates video from text prompts, still images, or both combined
- Followed Runway's earlier Gen-1 video-to-video style transfer model
- Integrated into Runway's broader suite of video editing and VFX tools
- Produces short clips typically a few seconds in length per generation
- Offered as a closed, hosted service under subscription pricing
- Not released with publicly available model weights
- Superseded by Runway's later Gen-3 model