OpenAI has released Sora 2, an updated version of its video generation model, and introduced a companion mobile app. The company said Sora 2 improves physical accuracy in generated motion and interactions, supporting activities such as sports routines and complex body dynamics. It is also designed to create longer, more coherent sequences without requiring separate shot-by-shot generations.
A major change in Sora 2 is that it generates synchronized audio alongside video. The model can produce dialogue in multiple languages, sound effects, and ambient soundscapes. Another new feature is Cameo, which allows users to insert digital versions of themselves, pets, or objects into generated scenes. Cameo requires user permission and identity validation before likenesses can be used.
The Sora app functions as a social platform where all shared content is AI-generated but posted by users. It includes identity profiles, a following system, remix tools, and moderation features. The company said it will apply safety policies, including watermarking exports, content provenance systems, and limits on harmful or unauthorized uses of Cameo.
The Sora iOS app launches first in the United States and Canada through an invite-based system, with an Android version in development. OpenAI said the model will also be accessible via its existing web platform and through an API in the coming weeks.
For more information, visit the OpenAI website or watch the livestream below
