As demonstrated by popular models, such as Dall-E and Stable Diffusion, generative neural networks have had significant success in generating short videos.
Runway AI's Gen 2 Gen 2 Network can create short videos based on a text description of only a few words. After that, the AI algorithm will create a 3-second video demonstrating exactly this or something very similar.
The Runway AI neural network is no longer available to everyone. Those who wish to try the neural network in action must register on the developers' website and join the waiting list. Last year, Runway AI raised $50 million in funding.
The neural network was asked to recreate a video of a "shooting a desert landscape from a drone." Several other videos illustrate the model's strengths and weaknesses. "Some cars can fly away," Runway AI cautions.
Gen 2 is capable of changing already finished videos by applying masks, adding stylization, or even completely reconstructing the picture. Or from a very simple 3D model to make almost a cinematic image.
The present Gen 2 algorithm is a direct continuation of the previously created Gen 1 neural network. Testing of the algorithm with the participation of a small number of users began in February. Currently, tens of thousands of users are involved in this process.
If you notice an error, click it with the mouse and press CTRL + ENTER.