Site navigation

OpenAI Introduces Sora, Its Realistic Text-to-video Model

Thom Carter

,

openai introduces its text to video model sora
OpenAI, arguably the world’s prominent artificial intelligence company, has revealed its latest innovation: a text-to-video model named Sora.

The blog post introducing the AI model states Sora is able to “generate complex scenes with multiple characters, specific types of motion, and accurate details of the subject and background.”

The videos, which run for up to a minute long, can be created from text prompts or from existing still images. The model can also extend existing video footage or fill in missing frames.

One of the highlighted examples of Sora’s capabilities depicts a woman walking down a Tokyo street. The video is highly realistic in spite of subtle giveaways that it’s artificial—such as irregularities in the camera movement and visual inconsistencies such as pavement not naturally interacting with the road below.

OpenAI acknowledged in its post that the current model has weaknesses—one of which being that it “may struggle with precise descriptions of events that take place over time, like following a specific camera trajectory”—despite the overall realism that Sora can produce.

In a bid to advance the model, the company is now granting access to a number of visual artists, designers, and filmmakers so that they can provide feedback on Sora.

The model is also becoming available to red teamers who are proficient in misinformation, hateful content, and bias to adversarially test Sora ahead of it becoming publicly available at a later date.

OpenAI noted that, to help with the detection of misleading content—and in addition to using existing safety tools, such as text classifiers that reject prompts requesting celebrity likeness—it’s building a detection classifier that can tell when a video has been generated by Sora.


Recommended reading


Speaking on why Sora has been introduced ahead of its release, the company said that it’s “sharing our research progress early to start working with and getting feedback from people outside of OpenAI and to give the public a sense of what AI capabilities are on the horizon.”

Thom Carter

Staff Writer, DIGIT

Latest News

AI

Nvidia Launches Open Secure AI Alliance for AI Safety and Security

AI Business Recruitment

Nearly a Quarter of Orgs Reducing Entry-level Hiring Due to AI Automation

Business

Scottish Businesses Turn to Self-funding as Growth Confidence Dips in H2

Data Finance

Payment Leaders are Struggling to Get Real-time Data