OpenAI’s Sora arrives on iOS with AI video and ‘Cameos’
NewsThe invite-only app launches in the US and Canada, powered by the new Sora 2 model with audio and controls over your likeness.

OpenAI has launched Sora, an iOS app and social feed for creating and sharing AI-generated videos, powered by the company’s next‑gen Sora 2 model with built‑in audio support.
Release date and price
The Sora app is available to download on iOS now in the United States and Canada, with access granted by invite. The app is free to use at launch, with OpenAI indicating paid options may appear during periods of high demand. ChatGPT Pro subscribers get access to a higher‑quality “Sora 2 Pro” model via sora.com.
Platform | Region | Access | Cost at launch |
---|---|---|---|
iOS app | US, Canada | Invite‑only | Free; may charge for extra generations during high demand |
Web (sora.com) | Invitees | Account required | Free Sora 2; Pro tier for ChatGPT Pro users |
You can install the app from the App Store.
New features in Sora 2
Sora 2 focuses on more realistic motion and better adherence to physical rules, spanning styles from photorealistic to cinematic and anime. The model now generates synchronized audio—speech, ambient sound, and effects—alongside video, reducing the need for separate sound design.
The app’s headline feature is “Cameos,” which lets you add yourself or friends to generated scenes after a one‑time video and audio capture to learn your appearance and voice. You control who can use your cameo and can see every video featuring your likeness, including drafts before they’re published.
“Prior video models are overoptimistic […] In Sora 2, if a basketball player misses a shot, it will rebound off the backboard.”
OpenAI describes these physics and motion improvements, and outlines the new model, in its Sora 2 announcement.
Security and privacy
OpenAI is layering controls around identity and feed personalization. Cameos require a short verification capture and are opt‑in. Permissions determine who can generate with your likeness, and you can revoke access at any time. OpenAI says cameo owners can remove any video using their likeness, including drafts created by others.
The feed prioritizes people you follow or interact with and includes tools to adjust personalization. The app also prompts users about wellbeing. For families, parental controls via ChatGPT can limit infinite scroll, toggle algorithmic personalization, and manage who can send DMs to a teen account.
On provenance and safety, OpenAI is adding measures such as attached metadata and visible watermarks on downloads, and it restricts creating photorealistic likenesses of public figures or content without consent.
Availability and rollout
Access to the Sora social feed is invite‑only by design, intended to bring people in alongside friends. Those invited can also use Sora 2 on the web. OpenAI plans an API to let third‑party apps integrate Sora 2, and an Android version is in development.
With generative video, audio, and tighter identity controls, Sora pushes short‑form creation toward fully synthetic clips while testing whether a friend‑centric, invite‑only model can keep the experience manageable at scale.
Comments