Overview
Sora 2 is OpenAI’s video and audio generation model, announced on September 30, 2025. OpenAI described it as improving physical realism, visual fidelity and controllability over its earlier system. (openai.com)
Capabilities and uses
- Generates video from natural-language instructions or an image reference, with synchronized audio such as dialogue and sound effects.
- Supports creative expression and storytelling, including realistic and imaginative scenes. OpenAI also presented it as a step toward models that better simulate the physical world. (developers.openai.com)
- The former API supported 720 × 1280 portrait and 1280 × 720 landscape output for
sora-2, with initial clip lengths of 4, 8, 12, 16 or 20 seconds. These specifications apply to the former API, not the separatesora-2-promodel. (developers.openai.com)
Limitations and availability
Generated scenes can contain mistakes and may not follow every instruction consistently; shorter clips may follow instructions more reliably. Outputs also raise risks such as misleading media and nonconsensual use of a person’s likeness. (developers.openai.com)
The model is no longer generally available through the OpenAI API. OpenAI’s deprecation record lists September 24, 2026 as the shutdown date for sora-2 and the Videos API, with no one-to-one replacement listed. The separate Sora consumer product was discontinued on April 26, 2026. (developers.openai.com)
Pricing
OpenAI’s pricing page, checked November 18, 2026, lists sora-2 video generation at USD $0.10 per second of output video for standard processing and USD $0.05 per second for Batch processing, at 720 × 1280 or 1280 × 720. The page does not list a separate input-token price for this model. These are listed API rates, not evidence of current API availability; the API was retired as noted above. (developers.openai.com)
Official sources
- Sora 2 announcement
- Sora 2 system card
- Sora 2 API model page
- API pricing — checked November 18, 2026.
- API deprecations