Sora (OpenAI)

On February 15, 2024, OpenAI published the technical report “Video generation models as world simulators,” introducing Sora, a text-to-video model. The report described Sora as a diffusion transformer: videos are compressed into a latent space and then split into “spacetime patches” that act as transformer tokens, allowing it to generate videos and images of varying durations, aspect ratios and resolutions; its largest model could generate a minute of high-fidelity video. The report’s authors included Tim Brooks, Bill PeeblesBill PeeblesBill Peebles is an artificial intelligence researcher, first author of the diffusion transformer (DiT) paper, who led OpenAI’s video-generation model Sora and, after leaving OpenAI in 2026, co-founded an AI video startup for filmmakers.Open the full entry and Aditya Ramesh.⁠[1]

The report also acknowledged that Sora had numerous limitations as a simulator, for example failing to accurately model some physical interactions such as glass shattering.⁠[1]

Sora was not initially released to the public; it was made available to red teamers to assess potential harms in areas such as misinformation, hateful content and bias, and to a number of visual artists, designers and filmmakers to gather feedback from creative professionals. OpenAI said it was building detection tools, including a classifier to identify videos generated by Sora, and planned to include C2PA metadata if it deployed the model in an OpenAI product.⁠[2]

Sora’s architecture drew on diffusion transformer (DiT) research, a project Saining Xie began in June 2022 with Bill Peebles, then interning at MetaMetaMeta (Meta Platforms, Inc.) is a publicly listed American technology company, renamed from Facebook in 2021, that operates social communication, digital advertising, virtual reality, smart glasses and artificial intelligence businesses.Open the full entry’s AI research lab, who later became co-lead of Sora at OpenAI.⁠[3]

On December 9, 2024, OpenAI released Sora as a standalone product at sora.com for ChatGPT Plus and Pro users, powered by Sora Turbo, a new version significantly faster than the model previewed in February. Users could generate videos up to 1080p and up to 20 seconds long in widescreen, vertical or square formats; Plus users could generate up to 50 videos a month at 480p or fewer at 720p, while the Pro plan offered 10 times the usage, higher resolutions and longer durations.⁠[4]

The release supported prompting with text, images and video, let users extend, remix and blend their own assets, and offered a Storyboard tool for specifying inputs frame by frame; the platform also had Featured and Recent feeds of community creations. Sora was not available to users under 18 or in the United Kingdom, Switzerland and the European Economic Area, and was not included in ChatGPT Team, Enterprise or Edu plans.⁠[4]

On September 30, 2025, OpenAI released Sora 2. OpenAI said the new model was more physically accurate than earlier models—for example, a missed basketball shot would rebound off the backboard rather than teleport into the hoop—and could generate highly realistic background soundscapes, speech and sound effects.⁠[5]

OpenAI simultaneously launched “Sora,” a social iOS app powered by Sora 2, initially rolled out by invitation in the United States and Canada. After a short recording to verify identity and capture their likeness, users could insert themselves into Sora-generated scenes; the feature, called “Cameo,” let users set who could use their likeness to only themselves, people they approved, mutuals or everyone. Sora 2 was initially free, ChatGPT Pro users could use the experimental, higher-quality Sora 2 Pro model on sora.com, and OpenAI said it planned to release Sora 2 in the API.⁠[5][6]

The Sora app rose to No. 3 overall on the U.S. App Store on its first day and reached No. 1 by October 3. Bill Peebles, head of Sora, announced on X that the app had passed one million downloads in under five days.⁠[7]

At launch, rightsholders’ works could appear in user-generated videos unless they explicitly opted out, an approach that placed the enforcement burden on rightsholders. On October 6, Motion Picture Association chairman and CEO Charles Rivkin issued a statement saying videos infringing members’ films, shows and characters had proliferated on Sora and across social media, that preventing infringement was OpenAI’s responsibility rather than rightsholders’, and that OpenAI needed to “take immediate and decisive action.”⁠[6][8]

In a blog post titled “Sora update #1,” OpenAI CEO Sam AltmanSam AltmanSam Altman is an entrepreneur and investor, and the co-founder and chief executive officer of OpenAI. He previously served as president of Y Combinator and founded the location-based social service Loopt. He also helped found Tools for Humanity and Merge Labs, has invested in energy and life sciences ventures, and has driven public discussion of AI governance, employment, and wealth distribution.Open the full entry said the company would give rightsholders more granular control over the generation of characters, similar to the opt-in model for likeness, and would try sharing some revenue with rightsholders.⁠[9]

Advocacy groups, academics and experts warned that generating videos from almost any text prompt had led to nonconsensual images and realistic deepfakes. After objections from family estates and an actors’ union, OpenAI cracked down on videos depicting public figures such as Martin Luther King Jr., Michael Jackson and Mister Rogers doing outlandish things.⁠[10]

On November 4, 2025, the Sora app launched on Android in the United States, Canada, Japan, Korea, Taiwan, Thailand and Vietnam, retaining the Cameo feature; OpenAI also planned “character cameos” featuring pets and objects. Appfigures estimated about 470,000 Android downloads on the first day, about 296,000 of them in the United States.⁠[11][12]

On December 11, 2025, The Walt Disney Company and OpenAI announced a three-year licensing agreement making Disney Sora’s first major content licensing partner: Sora and ChatGPT Images would be able to generate short user-prompted videos using more than 200 animated, masked and creature characters from Disney, Marvel, Pixar and Star WarsStar WarsStar Wars is a space-fantasy film and entertainment franchise created by George Lucas and now owned by Lucasfilm, a subsidiary of The Walt Disney Company.Open the full entry, excluding talent likenesses and voices, and curated Sora-generated videos would appear on Disney+. Under the agreement Disney would make a $1 billion equity investment in OpenAI and receive warrants to purchase additional equity; the deal was subject to definitive agreements and board approvals, with the features expected in early 2026.⁠[13]

On March 24, 2026, OpenAI announced on social media that it was saying goodbye to the Sora app, stating, “What you made with Sora mattered, and we know this news is disappointing.” An OpenAI spokesperson said the company had decided to discontinue Sora in the consumer app and API, citing growing compute demand and a research focus on robotics.⁠[10][14]

After topping the U.S. App Store following its debut, interest in Sora faded: Appfigures data showed month-over-month declines in new installs and user spending into early 2026, with December 2025 downloads down 32 percent from November.⁠[14]

The same day, Disney exited its deal with OpenAI, including its plan to take a $1 billion stake. In a statement, Disney said it respected “OpenAI’s decision to exit the video generation business and to shift its priorities elsewhere” and would continue to engage with AI platforms.⁠[15][10]

On April 17, 2026, Sora lead Bill Peebles announced his departure from OpenAI. Sora had been losing an estimated $1 million per day in compute costs.⁠[16]

On April 26, 2026, Sora’s web and app experiences were discontinued; the OpenAI Help Center said the Sora API would be discontinued on September 24, 2026, and that after discontinuation and any final export window, data associated with users’ use of Sora would be permanently deleted.⁠[17][4]

In October 2026, OpenAI was reportedly considering reviving Sora and was in talks with a stealth startup developing AI video for filmmakers. The startup sought access to Sora to customize it and sell software and services on top of it; its name had not been disclosed, and its co-founders included Jeffrey Katzenberg, former Sora research lead Bill Peebles and former Dropbox CFO Sujay JaswaSujay JaswaSujay Jaswa is an investor and former chief financial officer of Dropbox who co-founded the technology and media holding and investment company WndrCo with Jeffrey Katzenberg and in 2026 co-founded an AI video startup for filmmakers.Open the full entry, with Katzenberg and Jaswa having previously co-founded the investment firm WndrCo.⁠[18][19][20]

It was unclear whether Sora’s model weights would be shared, but if a deal went ahead the startup would still receive deep enough access to substantially customize the model.⁠[18][19]

OpenAI: Official website: https://openai.com/ (opens in a new window)⁠[4]