Your Digital Twin is here: OpenAI releases Sora 2
Editor's note, October 27, 2026: This essay retains its October 2, 2025 launch-week perspective. OpenAI has since discontinued Sora's app and web experience, effective April 26, 2026. The possibilities below remain predictions made at launch, rather than a description of the service today. OpenAI's discontinuation notice.
This is the Fortnite skin economy meets TikTok virality, except the skin is you.
This week OpenAI released Sora 2 alongside a new social app. Its cameo feature lets you generate videos using your likeness and voice, or bring friends into a scene with their permission. You can create, share and remix those videos inside the app. OpenAI's launch announcement.
The model update matters, but the social experiment interests me more. Your likeness becomes something other people can create with. Appearing in their video no longer requires being there.
Image by OpenAI
At first, a cameo feels like a novelty: a deepfake for fun. I think of it as a digital twin in the social sense, a version of yourself that can circulate through other people's creations. The more interesting possibility is what could accumulate around it: a recognizable style, a history of appearances, a network of friends and remixes. The feature captures a likeness; the people using it could give that likeness a social life.
It doesn't take much imagination to see an economy forming around this. Friend cameos are the obvious starting point, whether for a laugh or a flex. I've already had one friend text me to "open up your cameo dude!!"
From there, creators and influencers might pay people to help develop more polished versions of their digital selves. Brands could invite fans to put those selves in a campaign: trying on new sneakers, wearing a season's drop or appearing in a music video. If those creations run on the brand's account as well as the fan's, the relationship gets more participatory. Fans become part of the material a brand publishes. These are possibilities, not an established market.
That participation could also change how people think about synthetic video. A likeness used without permission can feel violating. A likeness you deliberately share with friends can feel playful. OpenAI says users control who can use their likeness and can revoke access. Those controls matter to the distinction. OpenAI's launch announcement.
I can imagine "That wasn't me, that was my cameo" becoming an ordinary explanation. That prospect is exciting and unsettling. Familiarity with consensual cameos could make synthetic appearances feel normal without resolving the harm of impersonation or misuse.
The data question is bigger than the joke. Creations, remixes and interactions could reveal which styles people prefer, who they want to appear with and how they want to be seen. OpenAI describes a personalized feed shaped partly by whom users follow and interact with. That establishes a recommendation mechanism; it doesn't establish that every share trains the video model. OpenAI's launch announcement.
My read is that this combination of identity, creative choices and relationships could become a substantial advantage for the platform. How much it contributes to future model training is a separate question. The social graph itself could be valuable even before that question is answered.
OpenAI's bet, as I see it, is that AI identity needs a social space of its own. If it works, the company could operate both the model that generates the videos and the network through which these versions of ourselves circulate. That's the possibility worth watching: a platform learning from the ways we choose to perform ourselves, then helping us perform again.