OpenAI Launches Sora 2 Model for Video & Audio Generation

OpenAI has launched Sora 2, a major leap in video and audio generation that improves physical accuracy, realism, and controllability. Building on the first Sora model from 2024, Sora 2 can handle complex, physics-consistent actions, such as Olympic gymnastics, paddleboard flips, and basketball rebounds, without morphing errors. It introduces synchronized dialogue and sound effects, supporting cinematic, realistic, and anime styles. A key feature is “cameos,” allowing users to insert themselves or friends into generated videos with high visual and vocal fidelity. The new Sora iOS app enables creation, remixing, and discovery in a social feed designed to prioritize creativity over consumption, with wellbeing safeguards and parental controls. Initially free with generous limits, Sora 2 is rolling out in the U.S. and Canada.

OpenAI, Research, Release, Product News – Sora 2 is here – 30th September 2025