OpenAI Launches Sora, Its First AI Video Generation Model 

OpenAI has announced the release of its highly anticipated video generation model, Sora. 

First previewed to select users in February, Sora allows users to create high-definition video clips using text prompts. Sora also generates video clips based on still images and can lengthen or fill in existing videos. 

Sora is capable of generating videos with up to 1080p resolution that are up to 20 seconds long in widescreen, vertical, or square aspect ratios. “You can bring your own assets to extend, remix, and blend, or generate entirely new content from text,” the company wrote in an announcement. A new storyboard tool also allows users to precisely specify inputs for each frame. 

The company also released a new version of Sora, Sora Turbo, that it says is significantly faster than the model previewed in February. Sora Turbo will be a standalone product available to ChatGPT Plus and Pro users.

OpenAI CEO Sam Altman joined researchers Aditya Ramesh and Bill Peebles in a livestream launch for Sora on Monday. (Source: OpenAI)

Learning About the Physical World

OpenAI says Sora “serves as a foundation for AI that understands and simulates reality—an important step towards developing models that can interact with the physical world.” Interaction with the physical world has been challenging for some models, as shown by a recent scientific study asserting that LLMs often fail to generate an accurate representation of the real world. 

The company noted that Sora in its current iteration often generates “unrealistic physics and struggles with complex actions over long durations.” 

In Sora's livestream debut Monday, OpenAI CEO Sam Altman said AI video is an important endeavor to OpenAI for several reasons. The company views video as a medium that fosters a co-creative dynamic between AI and human users that goes beyond text-based interaction. He also said video is crucial to OpenAI’s AGI roadmap, as it provides a rich environment for AI to learn about the real world. 

“We started the Sora project at OpenAI to build AI systems that really deeply understand the world and its physics,” said Bill Peebles, OpenAI’s lead research scientist for Sora, in the livestream. “We're just getting started here. This early version of Sora will make mistakes---it's not perfect. But it's already at the point where we think it's going to be really useful for augmenting human creativity, and we can't wait to see what the world is going to make with Sora.”

Ethical Questions Remain About AI-generated Video

Concerns about AI-generated video remain and are centered on potential misuse, including the spread of misinformation through deepfakes, challenges in detecting manipulated content, ethical issues around consent and privacy, and the risk of amplifying biases present in training data. 

A prompt window on the Sora interface. (Source: OpenAI)

OpenAI says all Sora-generated videos come with C2PA metadata to identify and verify their origin, while touting other safeguards like default visible watermarks and an internal search tool for additional verification. The company also says it is preemptively blocking “particularly damaging forms of abuse” like child abuse and sexual deepfakes. 

OpenAI says this metered introduction will give society time to explore the possibilities of video generation technology and develop more safeguards to ensure responsible use, but some experts aren’t convinced of its safety. 

Nana Nwachukwu, an AI ethics and governance consultant at Finnish AI governance firm Saidot, said: "The release of Sora, after a year-long delay, is a landmark moment for AI, but it also reignites vital discussions about safety and the ethical implications of advanced technologies. 

“While Sora’s potential to transform industries is immense, it also lowers the barrier to creating dangerously deceptive content,” she said, noting how OpenAI’s current guardrails may not be enough. “We cannot be sure of how effective these measures are or if the noted risks are all that is to be expected from this system. Governance must evolve alongside the technology to monitor and manage these risks. Without continuous oversight and robust industry standards, the promise of innovation risks being overshadowed by the threat of misinformation and harm.”

Availability

Sora will be available to ChatGPT Plus subscribers at no additional cost, allowing them to generate up to 50 videos monthly at 480p or fewer videos at 720p. OpenAI says its Pro plan will include 10x more usage, higher resolutions, and longer durations. The company is planning more tailored pricing plans for next year, it said. 

As of Tuesday at noon Eastern, Sora was unavailable to try, even for current OpenAI subscribers. The website displayed this message: “We’re currently experiencing heavy traffic and have temporarily disabled Sora account creation. If you've never logged into Sora before, please check back again soon.”

Visit the official announcement to watch a video demonstration of Sora.