OpenAI has decided to put an end to Sora, its ambitious generative video application with artificial intelligenceJust months after its massive launch as a standalone app and social network, the decision comes after a summer of high visibility for the tool, millions of downloads, and top-tier deals with giants like Disney. However, it also comes amid a growing debate about the costs, risks, and true fit of this type of technology within the company's business.
The announcement, made public through Sora's official X account (formerly Twitter), confirms that Both the consumer app and the developer API will be shut down.OpenAI has thanked those who participated in the community and announced that in the coming days it will provide instructions for preserving the generated videos, as well as a detailed schedule for shutting down the service.
What was Sora and why did it become such a rapid phenomenon?
Sora was born as OpenAI's first major bet on a standalone application focused on short videowith an interface very similar to TikTok or Instagram Reels. Since its public launch in September, users have been able to write text instructions, insert themselves into movie scenes or everyday situations, and share the result in a social feed of AI-generated vertical clips.
In its first days available to the general public, the app It surpassed one million downloads in less than five daysIt even surpassed ChatGPT's initial launch in some metrics. It topped the photo and video charts on the iPhone App Store and became the centerpiece of OpenAI's attempt to fully enter the short-video advertising market, dominated by TikTok, YouTube Shorts, Instagram, and Facebook.
Much of the initial momentum was based on the possibility of remix videos from other users and in the ease with which it can generate striking scenes: from impossible landscapes to recreations of pop culture icons. Sam Altman, CEO of OpenAI, even publicly encouraged users to play with their own image within recognizable moments from film and television.
After that dazzling start, the service was overwhelmed by demand and OpenAI had to rapidly expand the infrastructure needed to respond to video requestsOnce the initial peak had passed and usage had stabilized, downloads began to decline and Sora lost its presence in the rankings, something that already raised internal doubts about its medium-term trajectory.
The technology behind Sora and its hidden cost
Beyond its social network facade, Sora was one of OpenAI's most technically sophisticated projectsThe platform combined diffusion models—the same principles that have driven the generation of realistic images—with Transformer-like architectures, capable of interpreting text and visual sequences as a single flow of information.
The creative process began with a The video was completely noisy, similar to the static on a television with no signal.Through successive mathematical iterations guided by user instruction, the system gradually eliminated noise and revealed a coherent scene, both from a visual and narrative point of view.
Unlike other models that treat video as a simple succession of images, Sora divided the content into small three-dimensional space-time fragmentsThese patches took into account not only the width and height of each frame, but also the passage of time, which allowed for continuity between shots, consistency of objects, and the logic of their interactions throughout the sequence.
The model's training relied on millions of videos with which I learned the basic rules of the physical worldHow water behaves, how shadows are projected onto moving surfaces, or what happens when an object is bitten and that action should leave a trace in subsequent frames. This ability to simulate real-world phenomena was seen as a key asset not only for entertainment but also for future robotics.
However, all that technical deployment had a downside: the massive consumption of computing resourcesEach video generated entailed a high cost in computing power and, therefore, in infrastructure expenses, especially due to the function of the video cardAt a time when major tech companies are racing to build increasingly powerful data centers and the demand for AI chips has skyrocketed, maintaining a video-intensive social network became a difficult effort to justify.
OpenAI's strategic shift: goodbye to consumer video, hello to enterprise and robotics
The closure of Sora is not understood as an isolated case, but as part of a broader reorientation within OpenAIAccording to reports from media outlets such as The Wall Street Journal and other business newspapers, the company has internally announced that it will dismantle several products based on its video models, including the clip generation features integrated into ChatGPT.
The move fits with the plan of prioritize productivity tools and software for businesses and developersOpenAI has been merging previously disparate products—such as its ChatGPT desktop application, its programming technology, and its browser—into a sort of unified “superapp,” with which it intends to simplify its catalog and align its teams around a clearer product vision.
At the same time, the company has underlined its intention to Strengthen research in real-world simulation and roboticsCompany spokespeople have explained that the team that worked on Sora will continue to use video generation, but internally, to train robots capable of operating in physical environments and solving practical tasks.
This shift comes in a context where OpenAI is considering a possible IPO starting in the last quarter of the year.Becoming a publicly traded company means depending less on large private funding rounds and more on recurring revenue from stable products, which encourages a focus on businesses with clearer profitability, such as services for companies or programming solutions.
In this scenario, maintaining high-profile projects with less pronounced economic impact—such as a generative video social network with high computing costs—is perceived as a luxury that is difficult to sustain. Hence, within the company itself, the closure of Sora is interpreted as an attempt to “Let go of ballast” and reduce distractions to focus on what can sustain long-term growth.
Disney, deals up in the air and a setback for a major media showcase
One of the elements that attracted the most attention regarding Sora was his alliance with Disney and other major intellectual propertiesAt the end of the year, the entertainment company announced an agreement to allow users to generate videos with more than 200 characters from franchises such as Marvel, Pixar, or Star Wars within the OpenAI app.
That pact included the possibility that Sora would become an official showcase for content inspired by the Disney catalogThis is especially appealing amidst the ongoing battle for user attention on short-video platforms. Furthermore, various media outlets reported a multi-million dollar investment by the entertainment giant in OpenAI, partly linked to this collaboration.
OpenAI's strategic shift and the demise of Sora have put that project on hold. Disney spokespeople have indicated that the agreement will no longer proceed under the terms initially planned and have expressed that they respect the technology company's decision to abandon the generative video business aimed at the general public.
Disney insists that They will continue to explore collaborations with artificial intelligence platforms....provided they allow them to reach fans while respecting intellectual property and creators' rights. The Sora case thus serves as a reminder of how quickly priorities can change in a sector as volatile as generative AI.
The shutdown also impacts other, smaller-scale deals—including those with studios, content creators, and advertising agencies—that had begun experimenting with the platform. Many of these players now see how A potentially powerful distribution and promotion channel vanishes just when they were beginning to understand its logic and its audience.
Controversies: copyright, deepfakes and “AI garbage”
Beyond the business aspects, Sora has been surrounded by controversies linked to the use of intellectual property and the creation of deepfakesFrom its first weeks, copyright holders and audiovisual industry associations warned that the app allowed users to generate videos with images of real people, protected characters, and scenes easily mistaken for professional material.
Screenwriters' organizations and actors' unions, especially strong in the United States, expressed their concern about the impact these tools could have on creative employment and on control over one's own imageThe ease with which a user could produce scenes where a public figure appeared doing or saying virtually anything opened up an intense debate about the limits of this type of application.
Analyses from different media outlets and experts began to refer to Sora as a potential generator of “AI garbage”Large volumes of low-quality, repetitive, or blatantly deceptive videos were mixed with legitimate creative content. This visual noise made it difficult to distinguish the authentic from the fabricated and fueled fears of a new wave of disinformation based on hyperrealistic clips.
Faced with increasing pressure, OpenAI introduced additional controls to limit the creation of certain contentFor example, the creation of videos featuring characters such as Michael Jackson, Martin Luther King Jr., or iconic figures of popular culture without authorization was restricted after legal representatives and heirs raised their complaints.
Even so, the debate about the misuse of generative video and the responsibilities of the platforms remained alive. For some analysts, Regulatory and reputational issues have also weighed on the balance when deciding the future of Sora, especially in markets like Europe, where authorities are already working on specific frameworks for generative AI and its associated risks.
Fierce competition and pressure to find a clear direction
The Sora case is set within a competitive landscape in which OpenAI is no longer playing alone or in an unassailable positionWhile the company diversified its offerings with consumer products, social networks, assistants, development tools, and experimental projects, rivals like Anthropic have focused on a limited number of clear lines of business, primarily models for programming and enterprise use.
This narrower focus has allowed Anthropic gaining ground among developers and corporate clientsto the point of surpassing OpenAI in market share in some programming-related segments. Meanwhile, Google has been pushing its Gemini models, leveraging the enormous user base that already uses its search engine and services daily.
In the business sector, data from consulting firms and investment funds suggests that OpenAI has lost some of its initial advantageWhile ChatGPT is dominating in terms of the number of inquiries from individual users, the big question is how many of them pay for the service and how profitable it is to sustain that volume of traffic in the face of competition.
At the same time, the capital markets are shifting their tone. Key AI hardware manufacturers, such as NVIDIA, have warned that Mega-rounds of funding without clear return requirements will not last forever.In this context, companies in the sector know that they will soon have to showcase solid products with concrete business models if they want to maintain the confidence of future shareholders.
From this perspective, Sora's closure is interpreted as a symptom that OpenAI is trying to refine its proposal and avoid spreading itself too thin.The company has quickly gone from being seen as an undisputed pioneer to facing rivals with more focused roadmaps. Reducing flashy but peripheral projects is, in this scenario, a way to try to regain clarity and focus.
What will happen now to users, developers, and the ecosystem?
In its public communications, OpenAI has insisted that will publish detailed guides so users can export and save their videos before the platform shuts down completely. The company says it is exploring ways to facilitate the migration of content to other formats or services, although it has not yet specified exact dates.
For those who used the Sora API in their own projects—from third-party applications to creative experiments or internal tools—the news implies quickly rethink their roadmapThe discontinuation of the developer interface will force developers to look for alternatives in other generative video providers or to do without that functionality if they cannot find a viable substitute.
In Europe and Spain, the direct impact may be smaller in terms of absolute volume, since Sora's penetration was still incipient compared to other platformsHowever, numerous animation studios, advertising agencies, and digital creators had begun experimenting with the app as a laboratory for ideas and prototypes, something that will now be interrupted.
Some experts in digital law suggest that The closure comes just as the Old Continent prepares for the entry into force of new rules on AI.Algorithmic transparency and synthetic content. Less public exposure in generative video could, to some extent, alleviate some regulatory scrutiny, at least in the short term.
Meanwhile, the OpenAI teams that worked on Sora will be integrated into lines of work considered most strategic, such as “agent” systems capable of operating on the user's computer to automate tasks, or robotics projects supported by advanced simulations. The technical knowledge accumulated in Sora, especially in physical simulation and temporal coherence, will be reused in these fields.
The closure of Sora symbolizes a new stage for OpenAI: The company is moving from a phase of constant releases and a certain "experimenting in all directions" to a phase where focus, profitability, and regulatory viability weigh more heavily in every decision. Whether this adjustment comes in time to regain ground against its competitors remains to be seen, but everything suggests that consumer-driven video will no longer be the brand's main showcase in the coming years.
