OpenAI's Sora — Text-to-Video AI from Announcement to Shutdown

Sora is the AI model OpenAI announced on 15 February 2024 for generating video from text instructions. At announcement it could produce videos up to a minute long while keeping visual quality and adherence to the prompt, and it was offered first only to red teamers testing for harms and to a number of visual artists, designers and filmmakers. The public launch came on 9 December 2024, when a faster version, Sora Turbo, opened to ChatGPT Plus and Pro subscribers, with Japan among the first places it was available.

Sora 2 and its app, released on 30 September 2025, spread fast thanks to a feature that put users and their friends into videos, but drew a dispute involving the Japanese government, Japan's content industry and the US actors' union over videos resembling anime characters and impersonations of the dead and of actors. On 24 March 2026 OpenAI announced that it was discontinuing Sora. In about two years the product ran through the whole cycle of promise and friction that video generation brought.

The February 2024 announcement

OpenAI framed Sora as part of teaching AI to understand and simulate the physical world in motion, with the aim of models that help solve problems requiring real-world interaction. A technical report released the same day, "Video generation models as world simulators", described training text-conditional diffusion models jointly on videos and images of variable durations and resolutions.

The announcement showed samples such as a woman walking down a neon-lit Tokyo street and woolly mammoths in a snowy meadow, but also listed weaknesses, including trouble simulating complex physics and interactions between objects and multiple characters. On safety, it said that once Sora was in a product, text classifiers would reject prompts requesting extreme violence, sexual content, hateful imagery, celebrity likeness or the IP of others, and that it planned to include C2PA metadata.

The film industry reacted quickly. According to The Hollywood Reporter, the filmmaker Tyler Perry put on hold an $800 million expansion of his Atlanta studio, four years in the planning, which would have added 12 soundstages; seeing Sora, he felt his productions might not need to travel to locations or build sets. As an employer and as an actor and filmmaker, he also raised the alarm: "There's got to be some sort of regulations in order to protect us. If not, I just don't see how we survive."

The December 2024 launch: Japan in the first wave

The December launch let users generate videos up to 1080p and 20 seconds long, in widescreen, vertical or square formats. Plus included up to 50 videos a month at 480p, among other allowances, and Pro ten times the usage. OpenAI said every Sora video would carry C2PA metadata identifying its origin plus a visible watermark by default, that it was blocking child sexual abuse material and sexual deepfakes, and that uploads of real people would be limited at launch. The image above is the icon of C2PA's provenance label, Content Credentials.

Sora 2 and the dispute of 2025

OpenAI said Sora 2 was more physically accurate and controllable, could generate synchronised dialogue and sound effects, and might be the "GPT‑3.5 moment for video", its own assessment. The iOS app launched alongside it was a social app built around recording a short one-time video and audio clip to verify your identity and capture your likeness, then dropping yourself into any scene. According to TechCrunch it rose to the top of the App Store charts and passed 1 million downloads in a week.

Problems surfaced at once. According to INTERNET Watch, users in Japan quickly flagged that it could generate characters from Japanese anime and games, and on 6 October the Motion Picture Association in the US issued a statement calling for action. In his blog post "Sora update #1", Sam Altman wrote that OpenAI would give rightsholders more granular control over the generation of characters, including the option not to allow it at all, and added: "we are struck by how deep the connection between users and Japanese content is!" TechCrunch reported that OpenAI moved its policy for rights holders from opt-out to opt-in.

The Japanese government stepped in. According to a government written answer of 28 November 2025, the Cabinet Office's Intellectual Property Strategy Headquarters asked OpenAI not to engage in copyright-infringing conduct, as minister Minoru Jōnai disclosed at a press conference on 10 October. The answer also refers to Article 16 of Japan's AI Promotion Act, passed earlier that year. On 27 October the Content Overseas Distribution Association (CODA), whose members include 36 companies and 11 organisations in anime, manga, games and other media (as of April 2025), asked OpenAI not to use its members' content for training without permission and to respond in good faith to infringement complaints about Sora 2 outputs.

Impersonation was the other flashpoint. According to TechCrunch, after users posted disrespectful videos of Martin Luther King Jr., Sora paused depictions of him in October. The actor Bryan Cranston, whose likeness had been used even though he had opted out, joined SAG-AFTRA, OpenAI and others in a joint statement; he thanked OpenAI for improving its guardrails and asked all companies involved to respect performers' right to manage replication of their voice and likeness (Tom's Guide).

Shutdown: 24 March 2026

According to Engadget, on 24 March 2026 OpenAI posted from the Sora account: "We're saying goodbye to Sora." An OpenAI spokesperson told Engadget: "As we focus and compute demand grows, the Sora research team continues to focus on world simulation research to advance robotics." Engadget noted that, at the start of 2026, Appfigures data suggested month-over-month declines in both new installs and user spending, and, citing The Hollywood Reporter, that Disney was exiting the deal it had signed with OpenAI at the end of the previous year and would not invest $1 billion. According to OpenAI's help centre, the web and app experiences were discontinued on 26 April 2026 and the API will be discontinued on 24 September 2026.

What it meant for society

Sora's two years lined up, in quick succession, the questions video generation brought into society: the future of production jobs and investment (Perry's paused expansion), training data and character copyright (the Japanese government and CODA, and the switch from opt-out to opt-in), use of a person's face and voice without consent (King's family, Cranston and SAG-AFTRA), and ways to label where footage comes from (C2PA metadata and watermarks). The fight over generative AI and actors' and writers' work is covered on the 2023 Hollywood strikes page, and real harm from fake video and audio on the deepfake fraud page.

The record also shows a compute-hungry consumer video app that, for all the attention it drew, did not last as a business. The wider story is on the AI timeline and the AI and society timeline, and the developer on the OpenAI and Sam Altman pages.