← Video
Initial research 2026-09-21OpenAI

Sora 2

Inspect the documented capabilities, access conditions and sources for your task.

Initial research: 2026-09-21. Check current terms before choosing.

Updated

The facts

Version
2 (sora-2, sora-2-pro) — DISCONTINUED
Price
$0.75 per second of video. Usage rate. Settings and included usage can change the charge.
API
✕Not supported
Open weights
✕Not supported
Commercial use
Do not start anything on this. OpenAI notified developers on 24 Mar 2026; no new API keys since 1 Jun 2026; rate limits reduced in the final months; no successor model in the recommended-replacement column.
Max Clip Length
?Not verified

Moot — the API is removed in three days.

Source for Max Clip Length (opens in a new tab)
Audio Generation
✓Supported

Sora 2 generated synchronised audio; it was also credited with the strongest physical-world simulation while available.

Source for Audio Generation (opens in a new tab)
Image To Video
✓Supported

Supported while available.

Source for Image To Video (opens in a new tab)
Character Consistency
?Not verified

Not assessable now.

Source for Character Consistency (opens in a new tab)

Worth knowing

deprecation

OpenAI closed the Sora web and app experiences on 26 April 2026 and will permanently delete user content after the cutoffs.

2026-03-28The Decoder
deprecation

OpenAI removes the Videos API and every sora-2 model on 24 September 2026, with no replacement model listed.

2026-03-24OpenAI

Commercial risk

Not a licensing problem — a discontinuation. The Sora app was shut on 26 April 2026 and the Videos API plus all sora-2 models are removed on 24 September 2026, three days from today. OpenAI's next-generation model is codenamed Spud; Sam Altman has said pre-training is complete and video generation will continue inside ChatGPT, but no Spud video product, price or date has been announced.

From the publisher

Automatic source checks. These do not change the verification date of the facts above.

Sources

Beyond feature support

Quality evidence.

How well a model performs depends on the task, version, and settings.

We have not published a reviewed quality evaluation for these versions yet. Feature support and price do not establish output quality.

Try the same small task with each option. Check accuracy, consistency, and how much editing you need to get a usable result.

Start a conversation.