Sora 2, OpenAI's new AI-powered video-generating tool: how it works and what precautions to take

OpenAI unveiled Sora 2, its latest advancement in audiovisual content generation using artificial intelligence. The app, which converts written descriptions into videos with movement, sound, and characters, reached one million downloads in less than five days on the App Store, consolidating its position as one of the most viral tools of the year.

With this new version, OpenAI goes a step beyond ChatGPT by offering a visual and participatory platform. Users can create clips up to one minute long, interact with other creators, and customize the results with their own image or that of their friends. However, the excitement also comes with warnings about privacy, image rights, and possible misuse.

The launch of Sora 2 marks a turning point in the relationship between artificial intelligence and digital creativity. Its ability to generate hyperrealistic videos presents new opportunities in entertainment, education, and communication, but also reopens the debate about the ethical limits of the use of personal data and audiovisual manipulation.

A tool that converts text into video

Sora 2 works with an artificial intelligence model that transforms text into high-quality videos, with camera movement, synchronized sound, and dynamic scenes. Simply enter a description—for example, “a girl running through a field of sunflowers at sunset”—and the app generates a complete clip, complete with realistic lighting, depth of field, and effects.

Unlike previous versions, Sora 2 can produce video and audio simultaneously, incorporating music, voices, and environmental effects. It also uses a visible watermark to identify AI-generated content and prevent fraudulent or deceptive use.

The system allows users to adjust video length, camera angles, and visual styles. It is also designed to function as a social network: users can scroll through a personalized feed, like, comment, and share videos created by others.

Cameo, the Most Controversial Feature

One of Sora 2's most striking features is Cameo, which allows users to record short clips with the front camera so that AI can create a digital version of the user's face, voice, and gestures. This virtual replica can be integrated into the generated videos, interacting with other characters or scenes.

However, this feature has raised concerns among experts and regulators about its potential use in the creation of deepfakes. In response, OpenAI asserted that Sora 2 automatically blocks depictions of public figures and that users retain full control over their cameos: they can decide who uses them, delete their digital version, and restrict their appearance at any time.

The company also emphasizes that it has incorporated security mechanisms aimed at preventing abuse, including an automated moderation system and human reviews to detect inappropriate content.

Safety and Parental Controls

OpenAI has designed the app with specific protections for teens and young adults. The Sora 2 feed is moderated to display only appropriate material, and interaction between adults and minors is restricted. Youth profiles are not visible to adult users, and direct messaging features are limited to prevent unwanted contact.

According to the company, the goal is for Sora 2 to be "a creative tool to use with friends and family," not an uncontrolled open social network. These measures aim to avoid the security and privacy issues that have plagued other platforms based on AI or user-generated content.

Availability and International Expansion

For now, Sora 2 is available by invitation only and exclusively for iOS devices (iPhone and iPad). The app is currently active in the United States and Canada, although OpenAI announced that it plans to gradually expand to other markets, with priority given to Latin America, Europe, and Asia.

The company did not specify specific dates for its global launch, but announced that it will open early registration for creators and developers interested in experimenting with the technology.