OpenAI has unveiled its latest innovation, a video-generation model named Sora. This cutting-edge text-to-video AI has the remarkable ability to convert textual directives into both realistic and imaginative visuals. Users are given the power to generate photorealistic videos, with a runtime of up to one minute, purely anchored on their instructions. 

What sets Sora apart is its prowess in creating intricate scenes incorporating multiple characters, specific movement types, and detailed subject matter and backgrounds. This AI marvel exhibits a firm grasp of the physical existence of objects, while skilfully interpreting props and bringing to life vibrant characters oozing with emotion. Its versatility extends to generating videos from still images or adding missing frames and extending current videos.

Among Sora's impressive portfolio are demonstrations representing including a video of a stylish girl walking around evening Tokyo, and a vintage SUV speeding down a mountain road. Although there are mistakes in all videos, such as the "walking" wheel of an SUV, or the girl's right leg smoothly turning into the left, the overall output is nothing short of awe-inspiring.

Currently, Sora is open up to 'red teamers' for an in-depth assessment and crucial feedback. However, OpenAI acknowledges the need to proactively tackle the possible ramifications of AI-produced photorealistic videos being misconstrued as authentic footage.

Source: The Verge, Open AI, Sam Altman X (formerly Twitter)