Runway introduces image-to-video generation for Gen-3 Alpha and teases Gen-3 Alpha Turbo

It seems Runway has been quite active on X lately, announcing updates to its video generation offerings. The company recently shared that it would update Gen-3 Alpha to support image-to-video generation. With this update, users can upload an image for the model to use as the first frame of video clips up to 11 seconds long. Optionally, the images can be accompanied by a text-based prompt to provide the model with additional guidance. According to Runway, the Image-to-Video update brings increased artistic control and improves the consistency of the generated clips. As a demonstration of Gen-3 Alpha's image-processing capabilities, the company followed the announcement with a thread including 10 generations.

Less than 48 hours later, Runway announced it had been training an enhanced version of Gen-3 Alpha, introduced as Turbo, which delivers the same groundbreaking performance but generates videos several times faster. The sample video demonstrates Gen-3 Alpha Turbo generating a 10-second video in 11 seconds. The full video, which demonstrates the speed of both models side by side, has Gen-3 Alpha Turbo generating and playing an astonishing five different 10-second videos in the time it takes for Gen-3 Alpha to create and play a single 10-second video. Runway plans to roll out Gen-3 Alpha Turbo over the next few days to free and paid users, who will benefit from the lower pricing.

Runway is one of the AI companies to come under fire for their unethical data collection practices after 404 Media reported it obtained a training data spreadsheet comprising links to popular YouTube channels, spanning media and entertainment companies, news outlets, influencers, and other individual creators. The media company tested the model using keywords and prompts based on the document. 404 Media posted some of the resulting outputs, which bear an uncanny resemblance to the original videos referenced by the prompts.