Farewell Sora: This Might Be Good News for Enterprise AI Video Generation

In late March, OpenAI announced the end of their Sora Text-to-Video generative AI application and the eventual discontinuation of the Sora API. To understand the opportunity that the Sora-free enterprise market has ahead, perhaps it is worth taking an unvarnished view of Sora itself. While other models had the capacity to generate video from text prompts, Sora became perceived de facto leader in AI video generation because of its rapid initial adoption by broad consumer audiences, passing 1 million downloads in under 5 days in late September 2025.

The Cameo feature was an immediate hit with consumers, allowing them to upload images and create their own avatars or “clones” to insert into mems and generated videos. Sora unleashed the imagination of users, the vast majority of whom had never created videos or animations before. Just by typing what they imagined, Sora delivered close-enough-to-real videos. Sora also followed the “Sam Altman Playbook for Consumer Excitement” that worked so well with the text-to-text generative model ChatGPT: supercharge model training by introducing millions of excited consumers willing to prompt anything and everything.

But by February 2026, the cost and the free-falling user numbers started to turn into red flags. According to mobile analytics firm Appsfigure, Sora app downloads had dropped 32% month-on-month in December 2025, with another plummet of 45% in January. All told, Appsfigure estimated that the cumulative total from launch to shut down for in-app purchase revenue amounted to about $2.1 million…total. Reporting from the Wall Street Journal estimated that Sora’s operating costs were about $1 million…daily.

Users were increasingly disappointed with the quality of videos. Bad actors were actively sharing ways to bypass safety guardrails. Deepfakes of public figures in violent, highly disrespectful and even explicit and sexualized images were generated and widely shared. In the initial Sora launch, media companies needed to “opt-out” to explicitly say they did not want their characters or intellectual property to be used in AI generations. But after weeks of videos circulating of gross copyright violations, OpenAI relented and gave rightsholders a bit more control. Even then, it did little to stop the brand abuse.

In an October study by NewsGuard, Sora 2 produced videos advancing “provably false claims” 16 out of 20 times (80% of the time) when prompted to do so. One of the videos created depicted a Coca-Cola spokesperson announcing that the company would not sponsor the 2026 Super Bowl due to Bad Bunny performing in the Halftime show…a totally false claim. In fact, NewsGuard noted that while the model had guardrails against depicting known public figures, brands were NOT included in the protection.

In January 2026, Constellation Research, namely this author, was knee deep in a review of generative AI based applications focused on content generation. More specifically, I took to looking at the applications capable of generating images, graphics or video, but stopped short of analyzing individual foundation model performance. Sora was one of the apps considered for a new Shortlist that was released in February 2026. I looked at 22 solutions, many of which were applications that leveraged their own native model, but a good number would be better labeled as AI hubs, building robust creative applications around multiple models, giving creators choice based on creative need.

Among the criteria for inclusion in the Shortlist were things like security, privacy and guardrails expected for enterprise brand safety. Others tied back to the creative work at hand, looking for workflows, editing tools and tools built for collaboration. But perhaps most important would be the availability of tools to ensure fair use, eliminate plagiarism concerns, and the presence of hard and fast guardrails protecting against copyright infringement.

Sora didn’t make the shortlist. To be totally honest, it wasn’t even a close call.

My initial reaction in output review was that Sora, while fun for the casual consumer, was a potential liability within an enterprise content creation process. The output was entertaining but lacked quality and was often unusable for final production. Storyboarding and sketching out concepts might work, but for final cut, Sora would not be enough. The same issues that plague Dall-E, haunted Sora as the model consistently struggled with hands and feet, faceless crowds in backgrounds and distortion and noise in random places. There were just too many points of serious brand failure that could put marketing teams in the line of fire.

The death of Sora opens up a bigger conversation…dare I say forces us to look at the bigger picture for enterprises when it comes to where, how and why this type of AI is applied. This RIP notice should spark conversation within all creative, marketing and IT teams about minimum standards of AI governance and brand safety guardrails. We have graduated beyond the discussion of “should we use AI” and thanks to the hard lessons learned by the damage known as Sora, we should well understand what generative AI without guardrails, restraint or sense looks like.

For some, Sora generated video confirmed the need for more modern, AI powered brand intelligence and monitoring tools, capable of seeking out the good and the bad applications of a brand. For others, it has confirmed that the marketing tool kit is in need of a refresh and generative AI applications are top of mind for today’s Chief Marketing Officers. As consumers crave video driven experience, marketers must answer the call with expressions that maintain the brand identity and stay true to the organization’s governance policies.

Where should CMOs be looking when it comes to these generative AI applications? Did I mention there is a ShortList? All told, I landed on 7 solutions for image, graphics and video. You can view the full list here. Specific to video, here are 3 standouts…plus an honorable mention:

• Adobe Firefly: More than other solutions, Adobe Firefly stands out thanks to its intentional focus on commercially safe, enterprise ready content generation. The big “secret” about Adobe’s AI strategy is that the research teams behind Firefly are exceptional…and exceptionally picky. Just because they CAN generate, doesn’t mean they WILL. This is a company that intentionally chooses to NOT deliver a model that pumps out deepfakes. For brands, that strength of vision matters.
• Google Gemini AI Video Generator powered by Veo 3.1: Google is quickly racing forward with generative models like Nano Banana and Veo…and the output is terrific. A standout here is the amazing audio that can be generated, but the transparent and public safety guidelines and the adoption of SynthID should go a long way with enterprise creators.
• Leonardo.ai: Easy to use, easy to control is the goal for Leonardo’s AI Video Generator. From bringing motion to an image to building out short sequences that start with a sentence, Leonardo focuses on the creative process and production. Similar to the multi-model hub strategy from Adobe, Leonardo pulls power from Kling (also an exceptionally good AI Video model), Veo and more. Owned by Canva, Leonardo focuses on the creator, understanding that just like brushes, fonts and colors, video creators need the right tool for the right moment.
• HONORABLE MENTION: While not included in the Shortlist because it is part of a larger advertising platform, I would be remiss in not noting Amazon Ad’s Video Generator. Its delivering the video generation, turning static product images, brand assets and customer reviews into fully blown creative campaign assets through the “Creative Agent” within the late 2025 launched, Amazon Creative Studio. And, the videos are multi-scene, cinematic-quality video ads…generated in minutes.

So…marketers…let’s not spend too much time, if any, mourning Sora. It taught a valuable lesson that not all AI generation models are created equal…and not all applications are created for the enterprise. Sora sparked the imagination of consumers, who now, crave more video and understand that AI can generate worlds full of wonder. But for us…for the brand owners, creators and growth drivers, we aren’t looking for toys. We need tools. Sora was a creative diversion not meant for us to begin with. But the lessons learned from Sora should be remembered and safeguarded against.