Currently, the Microsoft-backed OpenAI has made Sora accessible to red teamers for studying critical areas for harm or risks, while having plans to extend access to several visual artists, designers, and filmmakers to gather their input on advancing the new AI model to provide optimal support for creative professionals
OpenAI unveils a new AI model, called Sora, to generate videos from text instructions. (Credit: Gerd Altmann from Pixabay)
OpenAI, a US-based artificial intelligence (AI) research and deployment company, has unveiled a new AI model called Sora that can generate videos based on text instructions.
The text-to-video AI model has been built to develop realistic and imaginative scenes that last up to 60 seconds. The output maintains visual quality while complying to the user’s prompt, said OpenAI.
Currently, the Microsoft-backed OpenAI has made Sora accessible to red teamers for studying critical areas for harm or risks.
The AI company will also make the tool available to several visual artists, designers, and filmmakers to get their input. This will be regarding how the new AI model can be advanced to provide optimal support for creative professionals.
OpenAI said the new AI model can create complex scenes with various characters, specific types of motion, and precise details of the subject as well as background.
In addition to comprehending the user’s request via the prompt, Sora is said to have knowledge of how those elements exist in the physical world.
Besides, Sora will produce multiple shots within a single video that maintain accuracy in depicting characters and visual style throughout.
Furthermore, the new AI model can take an existing still image and create a video from it by animating its contents with attention to small detail.
It can also take an existing video and extend it or fill in missing frames.
OpenAI stated: “Sora is a diffusion model, which generates a video by starting off with one that looks like static noise and gradually transforms it by removing the noise over many steps.
“Sora is capable of generating entire videos all at once or extending generated videos to make them longer. By giving the model foresight of many frames at a time, we’ve solved a challenging problem of making sure a subject stays the same even when it goes out of view temporarily.”
The AI company also revealed developing tools to help in spotting misleading content like a detection classifier that can determine when a video was generated by Sora.
OpenAI also intends to incorporate C2PA metadata in the future should it integrate the model into a product.