OpenAI launches Sora 2 with synchronized audio-video generation & a new standalone iOS app

OpenAI has released Sora 2, the next version of its video and audio generation model, arriving just 10 months after the original. Its main improvement is definitely the ability to generate audio that stays in sync with video, including spoken dialogue in multiple languages, bringing it closer to the quality seen in competing models such as Google’s Veo 3.

The update also improves how scenes follow physical dynamics and allows multi-shot instructions to keep environments consistent. It supports different visual styles like realistic, cinematic, and anime, and adds “cameos,” letting users upload short clips to place digital versions of themselves, friends, or objects into generated content.

Alongside the updated model, OpenAI has launched a new invite-only standalone Sora app for iOS, which functions more like a social network for sharing AI-generated videos, reminiscent of what was recently seen in Meta AI’s app. The new Sora app lets users create AI videos, remix them, and share them with cameo support. It includes common social features such as user profiles and customizable feeds, as well as natural language-based recommendations, well-being check-ins to help prevent overuse, teen usage limits, and parental controls powered by ChatGPT. Sora 2 is available on the web, with Pro subscribers gaining access to Sora 2 Pro and Sora 1 Turbo, and an API version is planned.