New Sora 2 Creates Incredible Videos

OpenAI has launched Sora 2, a powerful new AI model for generating realistic video and audio, alongside a companion app called Sora that functions as a short-form video platform in the style of TikTok.

Cameos

The app, now live in the US and Canada, allows users to create, share, and remix AI-generated videos featuring themselves and their friends through a feature called “cameos”.

OpenAI’s Move Into Social Apps

This marks OpenAI’s most significant step yet into the consumer social media space. While its ChatGPT platform has seen widespread adoption for productivity, the launch of the Sora app brings the company into direct competition with TikTok, Instagram Reels, and Meta’s new “Vibes” AI video feed.

Makes Short, Realistic Videos

The Sora app, available now for iOS on an invite-only basis, is designed around the idea of co-creation. For example, users can ask (prompt) Sora 2 to generate short, realistic videos with audio, then choose to insert themselves or others into the scenes using cameos. A one-time video and audio recording is required to activate this feature, verifying both identity and likeness.

Once uploaded, users can choose who may include their cameo in videos and revoke permission at any time. OpenAI says the app is designed to prioritise creativity and collaboration, rather than passive consumption. “We think a social app built around this ‘cameos’ feature is the best way to experience the magic of Sora 2,” the company stated in its launch blog.

What Can Sora 2 Do?

Sora 2 builds on the original Sora model released in February 2024, offering major improvements in physical realism and control. For example, the system now better obeys the laws of physics and can simulate complex multi-step actions, such as backflips, Olympic gymnastics, and buoyancy-based movements. It can also handle failure more realistically. For example, as OpenAI explains, “In Sora 2, if a basketball player misses a shot, it will rebound off the backboard,” thereby contrasting with earlier models that might teleport the ball into the hoop to satisfy a prompt.

The model can also generate sound effects and dialogue that synchronise with the visuals, and it supports different styles such as cinematic or anime. Sora 2 is also better at keeping things the same from one part of a video to the next, so people, objects, and scenes look consistent and don’t suddenly change or disappear.

Crucially, Sora 2 allows direct injection of real-world people or objects into generated videos. This is the basis for the cameos feature, which lets users appear in AI videos with recognisable facial likenesses and voices after completing the verification process.

A Social App Built Around AI Creativity

Inside the Sora app, users can browse a TikTok-style vertical feed where all content is generated by Sora 2. The feed’s algorithm is powered by OpenAI’s own language models and incorporates user activity, location, and engagement signals. Users can opt out of sharing data from their ChatGPT history, and parental controls allow for the disabling of infinite scroll, personalisation, and direct messaging.

The Sora feed has a different set of priorities than many existing social platforms. For example, according to OpenAI, the app has been explicitly designed to maximise creativity over time spent. Videos from friends or creators users have interacted with are prioritised, and content is recommended based on what the system believes users may want to remix or be inspired by.

OpenAI is also keen to highlight the fact that many early testers inside the company described the experience as a new way to communicate, i.e. beyond texts, emojis, and even voice notes. As the team says, “We all had a blast with it. It kind of felt like a natural evolution of communication”.

Safety And Moderation

It should be noted that the app includes several layers of moderation and safety, particularly around the use of likenesses, minors, and synthetic content. Videos with cameos include visible watermarks and embedded provenance data using C2PA standards. Also, only users with explicit permission can create content featuring another person’s cameo, and content involving under-18s is subject to stricter controls.

OpenAI says that a dedicated moderation team will review flagged videos, especially in cases involving bullying or misuse of likeness, and that it has developed a “new class of recommender algorithms” that can be instructed via natural language and will regularly poll users about their wellbeing to inform feed adjustments.

Reassuringly, it seems that parents will also have access to management tools via ChatGPT’s parental control features, which allow them to restrict viewing time and disable certain app capabilities.

Growing Interest

Despite being invite-only, the Sora app quickly topped the iOS App Store in the US following launch. According to Appfigures, it saw around 164,000 downloads in its first two days, suggesting strong early interest in AI-generated social video.

For individual users, the draw is clear, i.e. Sora offers a fast, no-equipment way to create and share realistic, creative videos. For businesses, Sora 2 may open up new workflows for marketing, campaign testing, training content, and generating creative ideas, all using synthetic media that’s watermarked and traceable.

It’s no surprise, therefore, that the model is already being explored for potential commercial uses. With this in mind, ChatGPT Pro users can access a higher-quality version called Sora 2 Pro on sora.com, and an API is planned, which would allow integration with third-party tools and platforms.

However, the move into social territory has also brought new scrutiny. Deepfake concerns, synthetic voice misuse, and intellectual property violations are all emerging challenges. OpenAI has already faced criticism from rights holders about the model’s ability to generate copyrighted characters or likenesses without permission. CEO Sam Altman recently responded on X (formerly Twitter), saying the company will move towards an opt-in model for characters and explore revenue sharing with rights holders.

Competitors And The AI Market

By launching its own social platform, OpenAI is taking a risk by entering a space dominated by highly entrenched players. That said, even though Meta’s “Vibes” feed, launched a week earlier, performs similar functions inside its Meta AI app, Sora goes further by integrating first-party video generation with a standalone app, a social graph, and content provenance controls.

Google is also moving rapidly in this space with Veo 3, which may bring advanced video generation tools to YouTube. However, OpenAI’s decision to bundle generation, remixing, social sharing, and personalised feeds into one app gives it a distinct early advantage.

Some would say, however, that the bigger prize here is data. For example, with Sora, OpenAI can gain access to a stream of video-based user interactions that could help refine future models, and reinforce its competitive position in the multimodal AI arms race.

Criticisms And Limitations

Although OpenAI has taken steps to address known risks, the Sora app is not without controversy. For example, some experts have raised concerns about likeness abuse, moderation gaps, and legal grey areas around synthetic content. Others are questioning whether the app’s feed algorithm, even if not optimised for time spent, could still encourage addictive behaviour.

Questions have also been raised about monetisation. For example, OpenAI says the app is free for now and that the only current revenue plan is charging for extra video generations during high demand. However, given the costs involved, a longer-term model will be needed. Sam Altman has indicated that monetisation will be transparent and designed to align with user wellbeing, but the details remain unclear.

As with other AI content tools, regulatory challenges will likely follow. For example, EU regulators, privacy watchdogs, and copyright bodies are expected to monitor Sora closely, especially if it expands into new regions or gains widespread use among teens and creators.

What Does This Mean For Your Business?

What happens next will depend on how well OpenAI manages the tensions between innovation, safety, and long-term sustainability. Sora has been built to showcase Sora 2’s technical strengths, but its future will depend just as much on how responsibly it operates as a platform. The company’s decision to emphasise provenance, permissions, and moderation from the outset suggests it has learned from previous criticism, but the real tests will come as usage scales and more complex edge cases arise.

For UK businesses, Sora introduces a fast, low-friction way to prototype video content, develop ideas, and test audience engagement using synthetic media. Creative agencies, marketers, educators and internal comms teams may all find practical value in exploring what this kind of AI generation can offer. That said, legal departments, compliance leads and brand managers will need to engage early. Controls around likeness, IP, and data use must be fully understood before models like Sora 2 are deployed commercially, especially in regulated sectors.

Broadly speaking, this release positions OpenAI not just as a model developer but as a direct provider of consumer experiences. This puts it on a collision course with existing social media and content platforms, and means its moderation policies, data practices and monetisation plans will face far more public and regulatory scrutiny. If OpenAI can maintain trust while building momentum, Sora could redefine how people create and interact with video. If not, it risks repeating the mistakes of the platforms it seeks to improve upon.