2 min read
OpenAI this week unveiled Sora 2, its next-generation video and audio model, alongside a companion social app simply called Sora.
The launch positions the company to push AI-generated short video directly into mainstream social feeds. The rollout is currently invite-only on iOS and will expand slowly across the U.S. and Canada as OpenAI manages enormous compute demands and safety controls.
Sora 2 is here. pic.twitter.com/hy95wDM5nB
— OpenAI (@OpenAI) September 30, 2025
What OpenAI Shipped and How You Use It
Sora is designed as a TikTok-style vertical feed of 10-second clips created with Sora 2. Each clip can include synchronized audio, and users can remix existing generations or create new ones from text prompts and scene controls.
The app’s standout capability is Cameo. After you complete a short identity check and provide a video or audio sample, the app inserts your likeness into AI-generated scenes. Creators strictly control Cameos, deciding who may use them and revoking access to any video that features their likeness.
Claims About Realism and Continuity
OpenAI highlighted that Sora 2 makes major strides in continuity, physical realism, and lip-synced audio. The company claims it is significantly better at “obeying the laws of physics” than prior models, reducing glitches such as warped movements or vanishing objects.
Early demos presented during the livestream appeared more coherent, though experts caution that Sora 2 still struggles with complex actions and multi-shot sequences.
Copyright Opt-Out and Hollywood’s Early Response
Unlike models that default to excluding copyrighted content, OpenAI has chosen an opt-out system, requiring rights holders to request exclusion if they don’t want their work used. Disney has already exercised this option, underscoring Hollywood’s concerns about AI’s use of copyrighted materials.
This move could set off a wave of similar decisions across the industry, as the app has the ability to remix or echo recognizable film and TV assets unless explicitly restricted.
