We’ve all been messing around with text and photo AI generation for a while now. This company wants to be the first to bring text-to-video AI generation to the masses.
Today, AI startup Runway announced its second-generation generative video model called Gen-2. While the first model took existing video and made changes or additions to it, the new model goes further, enabling users to input a photo or just text and get a video for the output.
The company says in its research report that its first-generation model, which used existing footage, would “require expensive re-training for every input or rely on error-prone propagation of image edits across frames.” The second-generation model, which can generate video off of text or still image input, is faster, more efficient, and offers users “fine-grained control over output characteristics, customization based on a few reference images, and a strong user preference towards results by our model.”
You can check out some examples of videos that the second-generation model created below:
The videos are obviously not photorealistic at this point but you can imagine the creative implications for anyone who wants to make a video but does not, and does not care to, know how to do so traditionally. The same shift is already occurring on the photo side with tools like DALL-E and Midjourney, where people are creating images and photographs without ever taking a picture or painting themselves.
As reported by The Verge, a spokesperson for Runway says that it will be rolling out general access to users in the coming weeks, but that access is still limited for now. Anyone who wants to try the tool out will need to sign up through the company’s Discord.
Generative AI is all the rage right now. While Runway is focused on text-to-video, the biggest example of the technology receiving adoption is Microsoft which just announced it is bringing generative AI tools to its productivity suite.
