OpenAI has unveiled Sora 2, a major upgrade to its AI video generation model, introducing realistic physics simulations, integrated high-quality audio, and a new social app for collaborative creation.
The announcement, made on September 30, 2025, marks a significant advancement from the original Sora model released last year, which faced criticism for inconsistencies in motion and visuals. Sora 2 addresses these issues by generating videos that adhere more closely to real-world physics, such as natural object trajectories and interactions, according to OpenAI’s official blog post.
Videos produced by the model now include synchronized audio elements, including ambient sounds, dialogue, and effects, eliminating the need for post-production dubbing. Demos showcased clips at 720p resolution and 30 frames per second, typically lasting 5 to 10 seconds, with options for photorealistic, cinematic or stylized formats like anime.
A standout feature of Sora 2 is “Cameo,” which allows users to record a short video and voice sample to insert themselves or others into generated scenes. Privacy controls let users manage access, view all usages, and revoke permissions at any time, OpenAI said. The company emphasized safeguards, including restrictions on deepfakes of public figures without consent and age-based limits for minors.
Complementing the model is the Sora app, available on iOS and the web, functioning as a social platform tailored for AI video sharing and remixing. Unlike traditional feeds, it prioritizes personalized recommendations based on user prompts, such as “more funny pet videos,” and fosters community collaboration without promoting passive consumption.
Access is currently limited to invite-only users in the U.S. and Canada, with free tiers offering generous generation quotas and a Pro option for enhanced features. An API for developers is slated for release soon, with global expansion planned.
OpenAI described the launch as a step toward “democratizing video creation,” potentially benefiting filmmakers, educators, and casual creators. However, the company reiterated its commitment to ethical AI, incorporating watermarking and content moderation to mitigate misuse.
For more details, you can read the original announcement on OpenAI’s blog – https://openai.com/index/sora-2/
