OpenAI has unveiled Sora 2, a major advance in AI video generation that emphasises realism, controllability, and synchronised audio.
The model introduces features that address long-standing weaknesses of generative video systems — from distorted physics to mismatched sound — making Sora 2 the company’s most sophisticated video and audio generator to date.
Sora 2 is here. pic.twitter.com/hy95wDM5nB
— OpenAI (@OpenAI) September 30, 2025
Alongside the model, OpenAI is launching a companion social app that allows users to generate and share short AI-produced clips in a TikTok-style feed.
The Sora 2 model represents a clear step forward over its predecessor. While the original Sora impressed with short, text-to-video outputs, it often struggled with consistent motion, scene stability, and sound alignment.
Sora 2 directly targets these issues. It integrates synchronised dialogue and sound effects, maintains consistent world states across sequences, and better reflects the physical laws that govern motion. For example, generated clips of athletes, vehicles, or natural environments now demonstrate fluid dynamics that align more closely with reality.
Features, Realism & Controls
At the core of Sora 2 is an emphasis on controllability. Users have greater influence over how characters move, how environments evolve, and how sound matches visuals. This makes the model not just a novelty tool but a potential asset for content creators, educators, and entertainment pipelines.
The system’s physics-aware design allows more accurate simulation of cause and effect: a ball bounces with proper trajectory, shadows fall consistently with lighting, and water responds realistically to interaction. Combined with synchronised dialogue and ambient sound effects, the outputs feel more polished and production-ready than prior iterations.
This is the Sora app, powered by Sora 2.
Inside the app, you can create, remix, and bring yourself or your friends into the scene through cameos—all within a customizable feed designed just for Sora videos.
See inside the Sora app👇 pic.twitter.com/GxzxdNZMYG
— OpenAI (@OpenAI) September 30, 2025
The companion app lets users generate up to 10-second videos in a swipe-based interface similar to TikTok. All content is AI-generated; uploads from device libraries are not supported at launch. A “Cameo” option enables verified users to lend their likeness to videos, creating new opportunities for personalisation — though it also raises questions around consent and moderation.
Access to Sora 2 is currently by invitation on iOS in the U.S. and Canada. Android support is planned but not yet confirmed. While usage is free within limits, paying ChatGPT Pro subscribers receive access to a higher-quality tier branded as Sora 2 Pro.
Implications & What Comes Next
By emphasising realism and control, Sora 2 places OpenAI in direct competition with not only TikTok and YouTube Shorts but also with emerging AI video platforms aiming to mainstream generative media. The ability to create believable, sound-synchronised content at scale could disrupt video production, influencer marketing, and entertainment workflows.
At the same time, challenges loom. Copyright enforcement, misuse of likenesses, and moderation of synthetic media remain unsettled. OpenAI’s safeguards block certain outputs, but ensuring safe use at global scale will test the company’s systems. Expansion beyond North America will bring further regulatory scrutiny.
Looking ahead, the roadmap points toward longer video clips, richer editing controls, and broader device support. If Sora 2 continues to improve on physical accuracy and controllability, it could set a new benchmark for what AI-generated video can achieve — edging closer to production-quality media created entirely from prompts.
