The generative-AI revolution is coming for video

The Future. Runway AI just released an AI generator that can turn text prompts into video clips. That’s a massive leap forward for the tech that has captured cultural attention for the past several months, demonstrating how fast the industry is evolving. While Runway’s tech may give the power of large-scale filmmaking to anyone with a computer, it may also massively complicate our ability to decipher between what is real and what is a deepfake.

AI killed the video star
Just a month after releasing its text-and-image-to-video generator, Gen 1, Runway is already dropping Gen 2.

• The system creates a three-second video simply from a text prompt (with the fewer words used, the better, surprisingly).
• Users can also upload an image as a “reference point” along with the text, but it’s not necessary like it was for Gen 1.
• The tool is via a waitlist, which people can sign up for on Runway’s private Discord channel. The company will add new batches of users weekly.

As of right now, the videos created by Gen 2 are silent, but Runway is in the early development of a system that can generate both images and sound.

No stability
Bloomberg tested out the tech and said it definitely does what it promises and is great at close-ups, but it does have a problem with things moving. Runway co-founder and CEO Cristóbal Valenzuela admitted, “You can generate a car chase, but sometimes the cars might fly away.” The team is still working through some kinks.

Nonetheless, Runway has beaten both Google and Meta to market on a vide-generating AI. Granted, Runway has been in the game for a while now — it worked on the original version of Stable Diffusion and raised $50 million last year to develop more AI tools.