OpenAI has introduced a new AI system, Sora, which is capable of creating videos based on text descriptions. The company’s blog post reports that Sora, by following custom prompts, can “create complex scenes with multiple characters, certain types of movement, and precise subject and background details.” The characters generated by Sora not only move, but also show vivid emotions; The visual quality of the video is maintained at a high level. Deep language understanding is integrated into Sora for more accurate interpretation of prompts.

However, Sora also has disadvantages, which the company is working to eliminate. For example, sometimes the system has problems with accurate physical modeling of objects, with the description of events, and with determining where is “right” and where is “left”. However, in general, Sora’s capabilities are impressive – it generates an entire video, not frame-by-frame, and the results of its work can be viewed here.

For security reasons, access to the new system is limited to a limited number of users. Currently, these are cybersecurity and disinformation specialists who study the risks of using Sora, as well as artists, designers, and filmmakers who need to evaluate the quality of video creation and identify flaws. They will be eliminated by OpenAI specialists, and later teachers, artists, and even politicians from all over the world will join the Sora test.

Літературний редактор.