OpenAI is introducing a social app where users can share AI videos with each other. The app is powered by the new Sora 2 model.
OpenAI has launched Sora 2, a new multimodal AI model that can generate realistic videos with matching audio from text prompts. At the same time, the company is introducing a new iOS app, Sora, where users can remix each other’s creations and create their own cameos via a one-time video recording. The app is currently available only in the US and Canada, with more countries to follow.
Video with Sound
Sora 2 is the successor to the original Sora model unveiled in early 2024. While that model was seen as the “GPT-1 moment” for video, OpenAI positions Sora 2 as a significant leap toward physically realistic simulations and better control over generated imagery. The model can portray complex movements and interactions, such as tumbling motions or flawed sports actions, and it accounts more accurately than earlier systems for physical principles like gravity, stiffness, and momentum.
Sora 2 also generates synchronized audio, including ambient sounds, speech, and sound effects. In addition, it can integrate elements from the real world into generated scenes—for example, by turning an existing video recording of a person into a digital cameo within a generated environment.
Social App
Alongside the model, OpenAI is launching a social app, Sora, centered on creating, sharing, and remixing videos using Sora 2. Users can add themselves or friends to generated scenes via the cameo feature. The app includes an algorithm that adapts based on preferences expressed in natural language.
The application is being rolled out in phases in the US and Canada. Users can sign up to request access. OpenAI emphasizes that the app is intended for creation, not endless consumption. For example, young users are limited in the number of videos they can watch per day, and there are parental control features via ChatGPT.