OpenAI has introduced the new video model Sora 2

OpenAI announced the launch of a new version of its video model — Sora 2. Compared to the first version, the updated neural network has improved realism, support for longer videos, and more accurate rendering of movements, faces, and lighting.

Sora 2 creates videos that are nearly indistinguishable from real ones, including complex scenes with multiple characters and dynamic environments. Additionally, the model has improved its handling of object "physics" — movements look more natural, and the interaction between objects and people in the frame has become more realistic.

Like the first version, Sora is available through the platform, but in CIS countries, the service is not yet officially available — a VPN is required to use it.

The launch of Sora 2 marks a new stage in the development of generative video and expands OpenAI's capabilities in the field of synthetic content.