OpenAI is taking another major step in AI video generation with the launch of Sora 2, a next-generation system capable of producing videos that look and sound remarkably real. The model improves upon its predecessor with enhanced physics simulation, enabling scenes that follow natural movement and logic rather than artificial approximations.
With Sora 2, users can generate intricate motion sequences—such as gymnasts flipping through the air or figure skaters performing precise spins—while synchronized dialogue and sound effects are automatically created to match. The system can even reconstruct real individuals and objects from uploaded videos, accurately reproducing their appearance and voice for insertion into new AI-generated scenes.
To complement the technology, OpenAI is rolling out a new Sora mobile app for iOS, offering users a platform to create, remix, and collaborate on AI-generated videos. Its cameo feature allows people to star in others’ projects, and co-creators retain the ability to remove their likeness from remixed clips. The app’s design prioritizes creative expression over endless scrolling, with moderation tools to encourage healthy engagement.
For now, Sora 2 is accessible via the web on an invitation basis, while the Sora iOS app will debut exclusively in the United States and Canada before expanding to other regions later on.