OpenAI launches AI system Sora, don't argue, 'video GPT' is still mine!

In a blog on its official website, OpenAI introduces its latest AI model - theVincent Video Model Sora, which generates raw video based on people's text descriptions.

In the demo video releasedSora Can generate clear and accurate roles, distant scenes, close-up scenes interlaced but the theme of the same video, and, has been able to generate the longest 60s of video at once. Watching these videos, AI automatically generate Hollywood movies according to the script, seems to have come to us.

As soon as it was released, startups related to AI and video on the social platform X couldn't sit still.

YouTube head netizen MrBeast responded to Sam Altman's tweet about the release of the Sora model by snarking, Sam, please don't make me homeless.

But OpenAI isn't stuck in the fancy video itself, for Sora Vincennes VideoThis is a milestone event on the road to AGI, says OpenAI.

According to OpenAI's official website, the goal of the Sora model is to "understand and simulate the physical world in motion". Currently, its highlight is the ability to generate videos up to one minute long, and, in that one minute, to ensure visual quality and adherence to user prompts. For example, Sora can generate videos of a series of complex scenes with multiple characters, specific types of motion, and precise details within a specific theme and context.

It understands not only what the user wants in a cue, but also how those things should exist in the physical world. Accordingly, Sora can accurately articulate and interpret cues based on its understanding of the cues, generating the right characters to express powerful emotions.

Of course, OpenAI also admits on its official website that Sora is far from perfect. It may have trouble accurately modeling the physical world in complex scenarios, and it may not be able to understand cause-and-effect relationships in some scenarios; for example, a person may take a bite out of a cookie, but afterward, there may be no bite marks on the cookie.

In addition, Sora may confuse spatial details such as 'left and right' in cue words, and may have difficulty accurately describing events over time.

OpenAI did not say when the Sora model would be made public, but said it would be shared with a limited number of outsiders. Throughout the day on February 15, local time, OpenAI CEO Sam Altman posted Sora-generated videos on his X social media account. He asked his followers to give him a cue word, and he posted the Sora-generated video for that cue word.

OpenAI says it will take some important security measures before applying Sora to OpenAI's products.

Regarding the significance of the Sora model, OpenAI concludes its official blog by stating, "Sora serves as the foundation for models that can understand and simulate the real world, and we believe that this capability will be an important milestone in the realization of AGI."

Obviously, OpenAI's goal is still the realization of AGI, and under this magnificent goal, its leading edge in gathering resources and its own ability once again reflects its comprehensive strength beyond that of ordinary startups.

In fact, this seemingly 'unconvincing' statement is inaccurate. the emergence of Sora has put an end to the product race in the 'video generation track', which was based on the previous generation of modeling technology. What starts now is a brand new technology competition, not just a product competition.