OpenAI Sora was a text-to-video model and social media app developed by OpenAI. Using artificial intelligence, the model generated short video clips based on text prompts and could extend existing short videos. It was first previewed in February 2024, publicly released in December 2024, and shut down in April 2026, with its API planned for discontinuation in September 2026. Sora's capabilities and rapid shutdown fueled discussions about generative AI's societal impact, including concerns over deepfakes and copyright.
Background
Before Sora, several other text-to-video models existed, including Meta's Make-A-Video, Runway's Gen-2, and Google Veo. OpenAI had previously released DALL-E 3, its third text-to-image model, in September 2023. Sora built on this lineage, adapting DALL-E 3's technology for video generation. The model was a diffusion transformer, a type of deep learning architecture that denoises 3D patches in latent space, then converts them to standard video via a decompressor. OpenAI trained Sora on publicly available and licensed copyrighted videos, but did not disclose the exact sources or quantities.
Initial Release
The team named Sora after the Japanese word for 'sky' to signify its 'limitless creative potential'. On February 15, 2024, OpenAI previewed Sora by releasing several high-definition video clips, including an SUV driving down a mountain road, an animation of a 'short fluffy monster' next to a candle, two people walking through Tokyo in the snow, and fake historical footage of the California gold rush. OpenAI stated it could generate videos up to one minute long. The company shared a technical report detailing the training methods. CEO Sam Altman also posted tweets responding to user prompts with Sora-generated videos.
As of December 9, 2024, OpenAI gradually made Sora available to the public for ChatGPT Pro and ChatGPT Plus users in the U.S. and Canada. Prior to this, a small 'red team' of experts in misinformation and bias conducted adversarial testing. OpenAI also shared Sora with a small group of creative professionals, including video makers and artists, for feedback. In February 2025, OpenAI announced plans to integrate Sora into ChatGPT, allowing users to generate videos directly from the chatbot.
Sora 2
Sora 2 was unveiled on September 30, 2025, alongside an iOS app, with an Android app following two months later. All videos generated by Sora 2 featured a visible, moving digital watermark to prevent misuse. The previous version also had a safety watermark. On October 7, 2025, 404 Media reported that third-party programs capable of removing the watermark had become prevalent. Many outlets, such as Wired, noted that the Sora 2 app was overtly similar to TikTok in style and features.
Discontinuation
On March 24, 2026, OpenAI announced on X that it was discontinuing Sora in both the mobile app and the API. The Sora app was shut down on April 26, 2026, while the API was planned to be shut down on September 24, 2026. OpenAI's partnership with Disney, which included a licensing agreement allowing Disney characters in Sora, was also ending. The decision prompted British technology news website The Register to label OpenAI a 'product-killer', following other tech companies like Google, Amazon Web Services, Broadcom, Cloud Software Group, and Netscape.
OpenAI did not provide a specific reason for discontinuing Sora in its shutdown notice. Reports linked the decision to computation shortages, cost pressures, and a broader shift toward core enterprise products. Following its public launch, Sora's worldwide users peaked at around a million before declining to fewer than 500,000, while the service cost an estimated $1 million per day to operate due to the computational demands of video generation.
Legal Regulation
In November 2024, an API key for Sora access was leaked by a group of testers on Hugging Face, who posted a manifesto protesting that Sora was used for 'art washing'. OpenAI revoked all access three hours after the leak and stated that 'hundreds of artists' had shaped development and that 'participation is voluntary'.
At launch, Sora 2 allowed copyrighted content by default unless copyright holders contacted OpenAI to restrict generation. On October 3, 2025, OpenAI stated that a future update would give copyright holders 'more granular control', but did not specify whether existing content would be removed. On October 6, the chairman of the MPA criticized OpenAI's approach to copyright with Sora 2.
On December 11, 2025, the Walt Disney Company announced a $1 billion investment in OpenAI, with a three-year licensing deal allowing Disney+ users to generate more than 200 copyrighted characters on Sora 2, including those from Disney Animation, Pixar, Marvel Studios, and Star Wars.
Capabilities and Limitations
The technology behind Sora was an adaptation of DALL-E 3's technology. According to OpenAI, Sora was a diffusion transformer, a denoising latent diffusion model with one transformer as its denoiser. Videos were generated in latent space by denoising 3D patches, then transformed to standard space by a video decompressor. Recaptioning was used to augment training data by creating detailed captions for videos.
OpenAI trained the model using publicly available videos and licensed copyrighted videos, but did not reveal the exact number or sources. Upon release, OpenAI acknowledged shortcomings, including limited capacity to simulate complex physics, understand causality, and differentiate left from right. OpenAI also stated that, in adherence to safety practices, Sora would restrict text prompts for sexual, violent, hateful, or celebrity imagery, as well as content featuring existing intellectual property.
Sora researcher Tim Brooks stated that the model learned how to create 3D graphics from its dataset alone, while fellow researcher Bill Peebles said that the model's ability to generate coherent scenes was a result of scaling up transformer architectures and training data.
Impact and Debate
Sora's release intensified public debate over deep learning and generative AI's potential for misuse, particularly regarding deepfakes. The visible watermark was a mitigation measure, but its removal by third-party tools highlighted challenges in content authenticity. The model's default use of copyrighted material, unless opt-out, drew criticism from copyright holders and industry groups, such as the MPA. Sora's discontinuation also raised questions about the sustainability of resource-intensive AI products, as operational costs and user decline contributed to its end.
Legacy
Sora's brief existence demonstrated both the technical potential and commercial challenges of text-to-video AI. Its diffusion transformer approach influenced subsequent research in video generation, while its shutdown underscored the economic realities of running such models at scale. The debates over copyright and watermarking that Sora sparked continue to shape policy discussions around AI content generation.