Sora is OpenAI's first generative Text-to-Video model that transforms text instructions into photorealistic videos. For the moment it is not open to the public but only to a selected group of security experts (who must test its reliability) and creators.
It is capable of generating videos of up to 60 seconds in HD quality starting from a simple text, exactly as happens for images with Dall-E 3 (also from OpenAI), Midjourney or Stable Diffusion. The company has not released technical specifications describing in more detail how the model works.