October 1st 2025
One year after Sora 1 redefined what was possible in video generation, OpenAI has unveiled Sora 2—its flagship system for video and audio generation—alongside the debut of the new Sora app. Streamed live just hours ago, the announcement positioned Sora 2 not only as a leap in model capability but also as a new platform for social creativity.
In February 2024, Sora 1 was described as the “GPT-1 moment” for video: the point where generative AI began showing object permanence and basic physical consistency. With Sora 2, OpenAI is claiming another step-change. The model improves on motion dynamics, body mechanics, and physical realism—enabling everything from gymnastic routines to wakeboard backflips with fluid believability.
The most notable feature is native audio generation. Unlike its predecessor, Sora 2 simultaneously produces video and sound: dialogue across languages, environmental soundscapes, and synchronized effects. The company emphasizes that this multimodal capability is general-purpose, allowing creators to script entire narratives in a single pass.
A standout innovation is Cameo, a feature that allows users to insert themselves—or their friends, pets, or even objects—into generated scenes. By analyzing a short clip, Sora 2 can replicate likenesses and identities inside any prompt, treating them as tokens within its world simulation framework. OpenAI is pitching Cameo as a new mode of communication, akin to the leap from text to emojis to short-form video.
To harness these capabilities, OpenAI launched the Sora app, a social platform where every piece of content is AI-generated but human-curated. The interface feels familiar—profiles, feeds, remix options—but the feed is powered by generative video. Users can create, remix, cameo friends, and participate in emergent trends.
Safety and identity control are built into the system. Cameos require liveness checks and explicit permission, while users retain rights to delete content generated with their likeness. Content is watermarked and provenance-tracked to ensure clear labeling beyond the platform. For younger audiences, usage limits and safeguards are in place to avoid endless scrolling.
The Sora app launches today on iOS in the US and Canada through an invite-based system. Each new user receives four invites to share with friends, emphasizing the social dimension. An Android version is in development. Beyond the app, Sora 2 will also power the existing web experience at sora.com, with creator-focused tools like storyboard editing on the horizon. An API is slated to follow in the coming weeks, opening the system to third-party integrations.
OpenAI framed Sora 2 as both a research milestone and a cultural experiment. By advancing world simulation, motion realism, and multimodal generation, the team sees it as a foundational capability on the road to AGI. But the launch also underlined another theme: joy. The demo emphasized playful creation—remixed perfume ads, dogs rendered in anime, and friends inserted into surreal stories.
For now, Sora 2 sits at the intersection of serious AI research and lighthearted digital culture, offering what OpenAI calls “the most powerful imagination engine ever built.”
