Open AI launches Sora, a revolutionary AI video tool | Technologies

While the ChatGPT and AI generative language models craze has not yet passed, OpenAI has just introduced an amazing and revolutionary video creation tool called Sora. With it, just give a description of what you want to see on the screen, and there it is, generated by artificial intelligence. Some are more advanced than others, sometimes they have that video game style that sets them apart from reality, but they are all amazing.

OpenAI CEO Sam Altman announced the launch on social network X, which was immediately inundated with new creations. Realistic, futuristic, crazy, cartoonish videos… The videos include all kinds of automated creations created using generative artificial intelligence. Sora is capable of creating entire videos at once or expanding created videos to make them longer.

Sora is capable of creating complex scenes with multiple characters, specific types of movements, and precise object and background details. According to OpenAI, the model understands not only what the user asked for in a query, but also how those things exist in the physical world. The model has a deep understanding of language, which allows it to accurately interpret signals and create compelling symbols that express powerful emotions, the company explains.

“Here is Sora, our model for video creation,” Altman wrote. “We’re offering access to a limited number of creators,” he added, before asking his followers to make suggestions for new videos in addition to the samples he already offered on his website.

The instructions may be more or less detailed. One example offered by OpenAI fits the following description: “An elegant woman walks down a Tokyo street filled with warm, bright neon and colorful city signs. She is wearing a black leather jacket, a long red dress, black boots and a black bag. She wears sunglasses and red lipstick. Walk confidently and carefree. The street is wet and shiny, creating a mirror effect of colored lights. There are a lot of pedestrians walking around.” And the result is surprising.

Another notes: “Movie trailer about the adventures of a 30-year-old astronaut wearing a red wool knit motorcycle helmet, blue sky, salt desert, cinematic style, shot on 35mm film, vibrant colors.”

In addition to being able to generate videos based solely on text instructions, the model is capable of taking an existing still image and creating a video from it, animating the contents of the image with precision and attention to fine detail. The model can also take existing video and enlarge it or fill in missing frames.

You can ask about content, style and give all kinds of instructions. Altman posts new videos requested by Twitter, proving the results are immediate. Sora can also create multiple frames in a single video, keeping the characters and visual style exactly the same.

“We teach AI to understand and model the physical world in motion, with the goal of training models that can help humans solve problems that require interaction in the real world,” OpenAI explains as it unveils its new text-to-video tool. “Sora can create videos up to one minute long while maintaining visual quality and accuracy of user instructions,” he adds.

The tool is currently available to so-called red teams. Members of these teams try to question the product or service, push it to its limits, test it and find fault with it, as if they were the enemy of the company. Here they have a specific task – to assess critical areas for potential damage or risks. They include experts in areas such as disinformation, hateful content and bias.

Open AI also provides access to a number of artists, designers and filmmakers so they can provide feedback on how to improve the model to make it more useful to creative professionals.

“We’re sharing our research early to start working with people outside of OpenAI to get their feedback, and to give the public insight into the AI ​​opportunities that are on the horizon,” the company explains.

Defects requiring polishing

The artificial intelligence firm itself admits that Sora still has some obvious shortcomings. You may find it difficult to accurately model the physics of a complex scene and may not be able to understand specific cases of cause and effect. As an example, he gives that a person can bite into a cookie, but then there may not be a bite mark left on the cookie.

The model may also confuse spatial details of the signal, such as confusing left and right, and may have trouble accurately describing events that occur over time, such as whether a camera follows a particular path.

OpenAI promises to take some precautions before making the tool available to the public. Among them is taking into account the instructions of the red teams. The company is also building tools to help detect misleading content, using detectors that can tell when a video was created by Sora. The company has also developed powerful image classifiers that are used to check the frames of all generated videos and ensure they comply with usage policies before displaying them to the user.

Additionally, you will reuse the security practices you create for your products that use DALL-E 3. For example, the text classifier will validate and reject text requests that violate your usage policies, such as those that require extreme violence, content images of a sexual nature. , hate images, images of celebrities, or the intellectual property of third parties.

“We will be reaching out to policymakers, educators and artists around the world to listen to their concerns and identify positive uses for this new technology. Despite extensive research and testing, we cannot predict all the beneficial ways people will use our technology or all the ways they will abuse it. That’s why we believe that learning from real-world use is a fundamental component of building and running increasingly secure AI systems over time,” concludes OpenAI.

You can follow El Pais Technology V Facebook And X or register here to receive our weekly newsletter.