Sora's Incredible Text-To-Video AI Proves Intensely Divisive

OpenAI has unveiled new software capable of creating remarkably realistic videos, but that has proven as controversial as it is impressive.

Is Sora a step too far?

OpenAI has dropped the pre-release version of its new text-to-video AI software, Sora, and the results are remarkable. The level of detail and realism is an order-of-magnitude better than previous AI video products.

In its current release, Sora can create videos up to 60 seconds long. Prompts can either be text instructions, or text plus an image. The output speaks for itself.

Sora uses two different approach to achieve the amazing levels of quality seen in some of these early examples. One is a "diffusion" model, much like the ones used in existing AI image generators like DALL-E. The other is so-called "transformer architecture", which puts the generated data together into a coherent sequence.

Mixed Reactions

While there's little doubt that Sora produces videos that are head and shoulders above the competition, the very realism of its output has proven highly divisive. On the one hand, there are those who recognize it as an incredible tool for creators to leverage.

Sora is good—very good—but it's not perfect (for example, like other AI image-generation packages, it still has problems with hands, among other things). But it's rightly being hailed as a step-change from previous AI video generators, because it's very convincing. Moreover, the technology will only continue to improve over time.

As New Scientist quotes Hany Farid, a researcher at the University of California: "As with other techniques in generative AI, there is no reason to believe that text-to-video will not continue to rapidly improve—moving us closer and closer to a time when it will be difficult to distinguish the fake from the real."

And therein lies the problem. "This technology, if combined with AI-powered voice cloning, could open up an entirely new front when it comes to creating deepfakes of people saying and doing things they never did."

The ability to create videos that are effectively indistinguishable from reality presents unprecedented possibilities for the propagation of misinformation, something that is already a problem with prior technology. It could also kill the movie industry at a stroke.

At this point, OpenAI does not have a timeline for releasing Sora more widely. For now, the organization is working on ensuring the safety of the new project, to whatever degree that is possible. For example, depictions of violence, sexual themes, and real celebrities and politicians, are not permitted.

What's clear, though, is that Sora marks a critical milestone for AI. We're entering a new era for generated content, and it won't be long before the line between reality and fiction becomes not just blurred, but almost eradicated.

Subscribe to our newsletter and follow us on X/Twitter.

Great! You’ve successfully signed up.

Welcome back! You've successfully signed in.

You've successfully subscribed to REX Wire.

Success! Check your email for magic link to sign-in.

Success! Your billing info has been updated.

Your billing was not updated.