Runway's New 3rd Generation AI Video Generator Wins High Praise

Runway, the AI ​​company known for its popular generative video tool, has announced its latest version, Runway Gen-3. The new model, which is still in alpha and not publicly available, was showcased in a set of example videos and appears to offer significant advances in terms of consistency, realism, and prompt adherence compared to the currently available Gen-2.

The videos generated were so realistic, especially those featuring human faces, that members of the AI ​​art community were quick to compare it to OpenAI's as-yet-unreleased but highly anticipated Sora.

“Even if they're handpicked, they're already better looking than Sora,” one Reddit user wrote in the most voted comment in the Runway Gen-3 discussion thread. “Sora has a stylized look and feel,” another user replied. “These figures actually look real and are the best I've seen so far.”

“If they'd shown me the generated characters I would have believed them to be real,” said another commenter on the AI ​​video subreddit, which has 66,000 members.

“These Runway GEN-3 clips are visually stunning to me – almost cinematic,” tweeted PZF, the anonymous AI filmmaker who also counts himself as a creative partner on Runway. “Smooth, understated (and natural in a good way), and believable.”

In addition to the Gen-3 video generator, Runway is also introducing a suite of fine-tuning tools, including more flexible image and camera controls.

“Gen-3 Alpha enhances Runway's text-to-video, image-to-video, and text-to-image tools, existing control modes like Motion Brush, advanced camera controls, and Director Mode, as well as upcoming tools that enable even more granular control over structure, style, and motion,” the company tweeted.

Runway claims that Gen-3 is a significant step towards realizing its ambitious goal of creating “general world models.” These models allow AI systems to build an internal representation of an environment and use it to simulate future events within that environment. This approach sets Runway apart from traditional techniques that focus on predicting the next likely frame in a given timeline.

Runway didn't reveal a specific release date for the Gen-3, but co-founder and CTO Anastasis Germanidis announced that the Gen-3 Alpha will be “available on Runway products soon,” which he hinted will include existing modes as well as “some new modes that were previously only available on the higher-performance base model.”

Runway's work in the AI ​​field began in 2021 when it worked with researchers from the University of Munich to build the first version of Stable Diffusion. Stability AI then stepped in to help offset the project's computing costs, turning it into a global phenomenon.

Since then, Runway has been a key player in the AI ​​video generation space, alongside competitors such as Pika Labs. That all changed when OpenAI introduced Sora, which surpassed the capabilities of existing models. Hollywood actor Ashton Kutcher recently caused controversy by saying that tools like Sora could be a major disruptor to TV and film production.

But as the world waits for Sora to be released to the public, new competitors have emerged, including Kuaishou's Kling and Luma AI's Dream Machine.

Kling, a Chinese video generator, can generate videos up to two minutes long at 30 frames per second in 1080p resolution, a significant improvement over existing models. The Chinese model is already available, but users must provide a Chinese phone number. Kuaishou said it plans to release a global version.

Meanwhile, Dream Machine is a free platform that converts text into dynamic videos, delivering results that far surpass Runway Gen-2 in terms of quality, consistency, and quick compliance. It does require a basic Google account, but because it's so popular, it takes a very long time for your generation to appear (if it even appears at all).

In the open source field, Stable Video Diffusion cannot produce comparable results, but it does provide a solid foundation for improvement and development. Vidu, another Chinese AI video generator developed by ShengShu Technology and Tsinghua University, uses a unique vision transformation model architecture called Universal Vision Transformer (U-ViT) to generate a 16-second video in 1080p resolution with a single click.

As for Pika Labs, no major updates have been released so its functionality remains on par with Runway Gen-2.

Decryption We reached out to Runway for more information on the release date and other details, but had not received a response as of this writing.

Editor: Ryan Ozawa.