Sora 2

What it is

Sora 2 is a second-generation video and audio generation model. OpenAI unveiled it on 30 September 2025, together with a new Sora iOS app.^[1][2]

Origins

Sora 2 succeeded the original Sora model from 2024.^[1] The new release arrived with a social app. The app launched by invitation for users in the United States and Canada and added social features, including cameos of users. Outlets compared it to TikTok, and an Android app followed about two months later.^[1]

Main features

The most visible feature was the cameo. Users could insert themselves and their friends into generated videos.^[2] Videos carried a visible, moving watermark. Within about a week, third-party watermark removers had appeared.^[1] A watermark is a visible mark placed over an image or video, usually to show where the content came from.

The API offered the models as sora-2 and sora-2-pro through a Videos API.^[4] Developers could therefore build on the model rather than only use the app.

Copyright and controls

Copyrighted characters were initially allowed unless rights holders opted out. After criticism from the Motion Picture Association and from Japan's Content Overseas Distribution Association, OpenAI promised more granular controls on 3 October 2025.^[1]

On 17 October 2025 OpenAI stopped deepfakes of Martin Luther King Jr. and let estates opt out.^[1]

Disney and the end of the service

On 11 December 2025 Disney announced a $1 billion investment in OpenAI and a three-year license for more than 200 characters to be used on Sora.^[1][3]

Usage was large at first. Worldwide users peaked near 1 million and then fell below 500,000. The service reportedly cost about $1 million a day to run.^[1]

On 24 March 2026 OpenAI announced that it was discontinuing Sora. The app shut down on 26 April 2026, and the Sora 2 API models were scheduled to shut down on 24 September 2026.^[1][4][5] The Disney partnership did not move forward after the shutdown.^[1]