OpenAI has just rolled out its new Sora 2 video generation model, packed with a host of advanced capabilities and enhancements. This artificial intelligence (AI) model debuts roughly a year after the original Sora was introduced. Users will be able to access this innovative AI video tool without cost through a new social media app, designed in the popular TikTok style. Currently, this iOS-exclusive app is available by invitation only, allowing individuals to create and share AI-generated videos featuring themselves or even leverage a “Cameo” function to make similar videos of others.
The Arrival of Sora 2 AI Model and Its New iOS App
In a detailed announcement, the San Francisco-based AI company elaborated on both the new AI model and its companion app. A significant leap forward is the inclusion of native audio generation, positioning Sora 2 competitively with offerings like Google’s Veo 3. OpenAI proudly describes Sora 2 as the “GPT‑3.5 moment for video,” emphasizing its ability to manage intricate movements and uphold realistic physics within generated videos.
OpenAI highlighted Sora 2’s arrival with a social media post, showcasing its advanced video capabilities and announcing the new app.
A compelling demonstration video featured a gymnast executing complex flips on a balance beam with remarkable realism, underscoring the model’s prowess. OpenAI also asserted that Sora 2 avoids morphing objects or distorting reality when processing prompts, which should significantly reduce visual “hallucinations” in the resulting videos.
The company also noted considerable improvements in output controllability, enabling Sora 2 to follow multi-shot instructions while maintaining consistent world states. Impressively, the video generation model excels in producing realistic, cinematic, and anime-style content.
Currently, the Sora 2 model is exclusively accessible via the Sora for iOS app. This invite-only application is restricted to users in Canada and the US. Its interface mirrors that of popular platforms such as TikTok, Instagram Reels, and YouTube Shorts, but with the key distinction that all its videos are generated by Sora 2.
A demonstration video within the Sora app revealed its user interface, showcasing how users can create, remix, and use the ‘Cameo’ feature.
It’s worth noting that Sora 2 video generation is presently offered for free to invited users. Once an invitation is secured, users can also generate videos directly from the Sora website.
Users have the option to upload a video of themselves, and the AI model, with its impressive character consistency, can then place their likeness into various other settings. These personalized videos can then be shared with followers, who can engage by watching, liking, commenting, and sharing. The innovative “Cameo” feature further allows users to select a character from an existing video and, using a text prompt, generate a new video featuring that character. Essentially, this enables the creation of personalized content, including “deepfakes” of friends.
OpenAI shared their excitement, stating, “We first started playing with this ‘upload yourself’ feature several months ago on the Sora team, and we all had a blast with it. It kind of felt like a natural evolution of communication—from text messages to emojis to voice notes to this.”
A crucial point is that identity verification is required before users can generate videos featuring themselves via the app. This measure prevents the unauthorized use of celebrity videos to create content. However, the “Cameo” feature introduces a potential bypass to some of these protective measures.
Interestingly, OpenAI is not alone in exploring the concept of vertically scrolling, AI-generated video apps. Recently, a major tech rival also launched its “Vibe feed” within its own AI app, offering users the ability to view short videos created using its in-house AI models.