Runway launches Gen-2 text-to-video to the public

Runway opened Gen-2, its text-to-video and image-to-video model, to the public through its website and an iOS app, ending a closed, waitlisted beta that had run through Discord since March. The model could generate short video clips from a text prompt alone, from a still image with a guiding text description, or by transferring the style of an image or prompt onto existing video footage — modes Runway packaged as text-to-video, image-to-video and stylisation.

Runway framed the release as a step toward removing the need for cameras and film crews in some kinds of production, and cited internal preference tests in which viewers favoured Gen-2’s output over comparable tools such as Stable Diffusion and Text2Live. Independent early reviews were considerably less enthusiastic about the results themselves: TechCrunch’s assessment, published the week of release, catalogued a low, near-slideshow framerate on the roughly four-second clips the model produced, visible graininess, and physics failures in which limbs and objects would merge or melt, concluding the tool was “more a novelty or toy than a genuinely useful tool in any video workflow.”

Gen-2 was nonetheless among the first text-to-video systems available to the general public rather than restricted to a research demo or closed beta, arriving as Runway competed with well-funded rivals racing to commercialise generative video before the underlying models were reliable enough for professional use. The gap between Runway’s stated ambitions and the reviewed quality of its output became a recurring pattern across the generative-video field over the following two years, as successive model generations from Runway and competitors steadily closed it.