AI 'Runway Gen 2' that can generate videos from text appears, the main battlefield of generative AI finally moves to video



AI startup Runway Research, which was involved in the development of the image AI `` Stable Diffusion '', released `` Gen-2 '', which generates video clips from text, on March 20, 2023.

Gen-2 by Runway

https://research.runwayml.com/gen2

Generative AI's Next Frontier Is Video - Bloomberg
https://www.bloomberg.com/news/articles/2023-03-20/generative-ai-s-next-frontier-is-video

Text-to-video AI inches closer as startup Runway announces new model - The Verge
https://www.theverge.com/2023/3/20/23648113/text-to-video-generative-ai-runway-ml-gen-2-model-access

You can see what kind of images can be created with 'Gen-2' announced this time from the following movies.



The predecessor of “Gen-2” is “Gen-1”, a generative AI that can synthesize new videos from existing videos, announced by Runway Research in February 2023. Using 'Gen-1', you can make a video of a box of sweets and a water bottle into a facility like a factory as shown below.



And with the 'Gen-2' announced this time, this function has been further developed, and it is now possible to generate images from text. For example, if you enter the text 'Drone image of desert landscape', a desolate landscape image will be output.



Complex instructions such as 'the afternoon sun peeking through a loft window in New York' are also possible. Meta's ``

Make A Video '' and Google's `` Imagen Video '' are leading the way in similar AI, but Runway is in a line with the generative AI of large companies in that it was developed by a team of only 45 people. is drawing



All of the demo clips are about 3 seconds long, the image quality is rough, and the frame rate is low. It's also not available to everyone, so far you'll have to register via Discord to be added to the waitlist. A Runway spokesperson said Gen-2 will be available to broader access in the coming weeks.

In addition, Runway CEO Cris Valenzuela told Bloomberg, 'Although the video generated by 'Gen-2' is silent, we are also researching sound generation with the aim of creating a system that can also generate sound in the future. I'm here,' he said.

Related Posts:

in Software,   Video, Posted by log1l_ks