OpenAI, which sprang to prominence last year owing to the success of ChatGPT, is expanding its artificial intelligence capabilities to video.

On Thursday, the business debuted Sora, its latest generative AI model. Sora functions similarly to OpenAI's image-generation AI tool, DALL-E. When a user enters in a chosen scenario, Sora will return a high-definition video clip. Sora can also create video clips based on still photographs, expand existing films, and fill in missing frames.

Following the success of chatbots and image generators in the consumer and commercial worlds, video might be the next frontier for generative AI. While the creative options will delight AI aficionados, the new technologies raise real worries about disinformation as key political elections approach across the world. Clarity, a machine learning business, reports that the number of AI-generated deep fakes has surged by 900%  year on year.

OpenAI's Sora competes with other video-generation AI tools, like Meta and Google's Lumiere (introduced in January). Other firms provide similar AI solutions, such as Stability AI's Stable Video Diffusion. Amazon's Create with Alexa concept creates short-form animated material for children based on prompts.

Sora is now limited to producing videos that are one minute or shorter. OpenAI, supported by Microsoft, aims to provide a greater range of AI models by including multimodality (text, picture, video production).

Sora had previously only been available to a restricted set of safety testers, or "red teamers," who examine the model for flaws such as disinformation and prejudice. The business hasn't published any public demos other than the 10 sample clips accessible on its website, and it says the accompanying technical paper will be issued later on Thursday.

OpenAI also stated that it is developing a "detection classifier" capable of detecting Sora-generated video snippets, as well as that it intends to incorporate metadata in its output that will aid in the identification of artificial intelligence material. during is the same type of metadata that Meta hopes to utilize to detect AI-generated photographs during election season.

Sora is a diffusion AI model that, like ChatGPT, employs the Transformer architecture, which was developed by Google researchers in a 2017 publication.

"Sora serves as a foundation for models that can understand and simulate the real world," OpenAI explained in their release.