OpenAI has officially announced Sora 2, its new audio and video generator, alongside a linked social app called Sora, allowing users to share and generate artificial intelligence (AI)-powered videos on a TikTok-style feed.
The app includes an “upload yourself” feature called cameos, which users can easily drop themselves into generated scenes.
To enable this, users are only required to upload a one-time video-and-audio recording for identity verification. Moreover, users can allow their friends to use their likeness in multi-person videos.
According to the ChatGPT manufacturer, Sora 2 enhances realism by better following the laws of physics.
Unlike previous models that morphed objects unrealistically, Sora 2 offers natural clips, similar to a basketball rebounding off the backboard rather than teleporting into the hoop.
Demo videos display beach volleyball, skateboarding, diving, and other exciting activities.
Initially, the Sora iOS app was introduced across Canada and the US, with plans for wider expansions in the near future.
While the app is invite-only, ChatGPT Pro subscribers can try the recently launched models.
Notably, videos generated by the Sora 2 Pro model can be shared in an algorithmic feed, similar to TikTok or Instagram Reels.
To personalize recommendations, OpenAI factors in user activity, location, engagement, and ChatGPT history, though users can easily disable it whenever they want.
Parental controls let guardians limit scrolling, block personalization, and manage messaging.
The app is currently free at release, with monetisation planned through paid video generation during peak demand.
However, experts cautioned that cameo features significantly increases risks of misuse, include a non-consensual and deceptive AI-generated videos.
