OpenAI Unleashes Realistic Text-to-Video ‘Sora’ AI


OpenAI has a new text-to-video generative AI model called Sora, and is launching the tool in limited access to red teamers for testing as well as selected filmmakers and artists.

Like its other generative AI offering ChatGPT, Sora generates responses based on user-submitted text prompts. It can create videos up to 60 seconds long, according to the company.

Red teamers are a tech and cybersecurity industry term for individuals hired to assess a product’s potential risks from an adversary’s perspective. OpenAI says Sora is now available to such groups to test its safety.

Sora-generated footage published by OpenAI suggests the tool is a substantial leap forward in realistic AI-powered video. Sora can understand “camera” positioning, suggesting it could mimic film industry camera operator movements when provided via text input. Users can write as little as a sentence or a paragraph to describe the video they want Sora to create.

But OpenAI warns that Sora isn’t perfect. “The current model has weaknesses,” the company says. “It may struggle with accurately simulating the physics of a complex scene, and may not understand specific instances of cause and effect.”

Sora could also make basic errors like the difference between left and right, OpenAI says. The tool’s camera imitation may also not be consistent over time. It may also spontaneously spawn new characters in random locations, as if they appeared by magic.

Some of the humans in Sora’s creations are also uncanny and bizarre, doing things like waving with a bent thumb or holding their hands in strange ways. Yes, AI still struggles to grasp the concept of human hands.

But any new photorealistic AI content generator raises concerns around potential biases, as well as the potential for increasingly convincing AI deepfakes.

Recommended by Our Editors

OpenAI says it will add provisions that block Sora from generating any content that mimics celebrity likenesses or generates hateful, extremely violent, or sexually explicit content. The company will also add metadata to videos marking Sora-made content as AI-generated, according to the announcement. It’s currently easy to remove this metadata, however.

OpenAI’s latest tool offers some impressive results, but any video-generation tools could pose an inherent risk to the livelihoods of production crew, camera operators, cinematographers, and others working in the film, television, and digital media industries. It could also speed up production timelines, potentially allowing previsualization teams or concept artists to create and realize more ideas faster.

Like Hollywood’s writers and actors, who went on strike last year in part due to concerns over AI tools, the film production crew union IATSE is set to begin its negotiations with the film producers in the Alliance of Motion Picture and Television Producers next month. It’s likely that generative AI, especially given the launch of new tools like Sora, will be a hot-button issue.

Get Our Best Stories!

Sign up for What’s New Now to get our top stories delivered to your inbox every morning.

This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of Use and Privacy Policy. You may unsubscribe from the newsletters at any time.


Leave a Comment