Sora 2.
OpenAI's video model — the production-grade reference for synthetic film.
What it is.
OpenAI's diffusion video model. Sora 2 shipped end of 2025 with 60-second clips, sound, and controllable cameras. Sora 2 Edit handles inpainting and continuation. Distributed through sora.com, ChatGPT, and the Sora API. The reference point every other video model is benchmarked against.
Where it fits.
Brand and marketing teams running ad iterations against media plans they cannot afford to shoot. Pre-production pipelines at studios that already license OpenAI for writers' rooms. Anywhere the budget for synthetic video is shifting from agencies to in-house operators.
- Best motion coherence and physics in production
- Native sound generation in Sora 2
- API access for programmatic pipelines
- Rights and likeness handling still evolving
- API quotas tight at launch for non-enterprise tier
Frequently asked.
Can Sora generate clips with my brand assets?
Yes via the Edit endpoint and via uploaded reference frames. Likeness use is gated behind verification.
What is the maximum clip length?
Sora 2 generates 60-second clips natively. Longer outputs are produced via continuation through Sora 2 Edit.
Are commercial rights cleared for outputs?
Yes for paid tiers. Free outputs carry a watermark and a non-commercial licence.
