Sora 2 Launches

OpenAI has unveiled Sora 2, its next-generation flagship video and audio generation model. If the previous model was at the 'GPT-1' stage of demonstrating the possibilities of video generation, Sora 2 is being evaluated as having entered the 'GPT-3.5' stage of understanding and simulating physical laws more deeply.

Sora 2 has achieved the following key advancements:

• Improved Physical Accuracy: It models complex dynamics such as buoyancy and rigidity, reproducing not only successful scenes but also physics-law-based 'mistakes.'
• Expanded Control and Style: It follows complex instructions and implements various styles such as cinematic and animated.
• Audio Synchronization: It generates highly realistic background sounds, voices, and sound effects along with the video.

Through the new 'Characters' feature, users can inject their own appearance and voice directly into the model. This allows users to feature themselves or friends in any generated environment with high fidelity.

To experience these features, a dedicated Sora iOS social app has also been launched. Within the app, users can create content, remix, and explore a customized feed. To prevent issues of addiction and excessive dopamine exposure, OpenAI plans to provide tools that allow users to directly control their feed.