Runway Unveils "Gen-2": The World's First Publicly Available Full Text-to-Video AI
Runway announced its second-generation model, “Gen-2,” capable of generating entirely new video from nothing but a text prompt or a single image. Whereas its predecessor Gen-1 was limited to restyling existing footage, Gen-2 requires no source video at all — and Runway positioned it as “the first publicly available text-to-video AI.”
Details
- Input methods: Can generate new video from text prompts alone, from an image alone, or from a combination of text and image
- Availability: Access was still limited at launch, requiring interested users to sign up via Discord, with Runway saying it would gradually widen general access over the following weeks
- Quality issues: Once released, users pointed out rough edges typical of generative video AI at the time — flickering outlines, movement that ignored the laws of physics, and unnatural blending of arms and legs
- Development background: Developed by Runway, which had also been involved in Stable Diffusion’s development, and backed by major investors including Google and Nvidia
What happened next
Gen-2’s arrival marked a turning point where video generation AI evolved from “editing existing footage” to “generating entirely new content from text.” At the end of June that year, the company secured an additional roughly $141 million in a Series C round from investors including Google, Nvidia, and Salesforce, laying the investment and technical groundwork for Gen-3 Alpha in 2024 and Gen-4 in 2025.