Sora 2 is the second-generation video generation model from OpenAI, featuring significant advancements in physical simulation and native audio integration. While the base model was initially unveiled in September 2025, the December 2025 update marked a major production milestone, introducing a dedicated mobile app and a landmark partnership with the Walt Disney Company that integrated over 200 licensed characters into the generation engine.
The model is built on an enhanced Diffusion Transformer (DiT) architecture that processes video and audio as unified spacetime patches. This architecture enables native synchronized audio, allowing the model to generate speech, sound effects, and ambient noise that align with visual motion in a single pass. Key features of Sora 2 include Cameos, which allow users to verify their identity and insert their own likeness into scenes, and the Remix tool for interactive editing and stylistic iteration of generated clips.