Sora was OpenAI's text-to-video system: a research model first, then a standalone product, then a consumer app built around the Sora 2 model. As of 2026-11-19 it can no longer be used. According to OpenAI, the Sora web and app experiences were discontinued on April 26, 2026, and the Sora API followed on September 24, 2026.
Where Sora came from
OpenAI introduced Sora on February 15, 2024, in a technical report titled "Video generation models as world simulators." The report describes a diffusion model that works on "spacetime patches" taken from compressed video, using a transformer architecture, a design OpenAI calls a diffusion transformer. Its main claim was breadth: one model that could generate videos and images of different durations, aspect ratios and resolutions, including widescreen 1920x1080 and vertical 1080x1920. OpenAI said its largest model could generate a minute of high-fidelity video. The report also states that model and implementation details were not included, so it works as a description of the approach and a qualitative evaluation, not a recipe.
The December 2024 product
On December 9, 2024, OpenAI moved Sora out of research preview with a faster variant called Sora Turbo, released as a standalone product at sora.com for ChatGPT Plus and Pro users. According to the announcement:
- Videos could reach 1080p resolution and 20 seconds as the product maximum, in widescreen, vertical or square formats. The Plus plan was capped at 480p and 720p (see the limits below).
- Users could bring their own assets to extend, remix or blend, or generate from text alone.
- A storyboard tool let users specify inputs for individual frames.
- Featured and Recent feeds showed community creations.
The plan limits were part of the announcement. Plus included up to 50 videos at 480p resolution, or fewer videos at 720p, each month. Pro included ten times more usage, higher resolutions and longer durations. Sora was not included with ChatGPT Team, Enterprise or Edu, was not available to people under 18, and at launch was offered wherever ChatGPT was available except the United Kingdom, Switzerland and the European Economic Area.
Limits and safeguards OpenAI described
OpenAI was direct about weaknesses. The launch post says the deployed version "often generates unrealistic physics and struggles with complex actions over long durations." On the safety side, it lists several measures:
- C2PA provenance metadata on all Sora-generated videos, plus visible watermarks by default.
- An internal search tool that uses technical attributes of generations to help verify whether content came from Sora.
- Blocks on child sexual abuse material and sexual deepfakes, with limits on uploads of people at launch.
The system card adds age gating to users 18 or older, restrictions on face uploads, and more conservative moderation thresholds for prompts and uploads involving minors. It says OpenAI worked with hundreds of visual artists, designers and filmmakers from more than 60 countries after the 2024 announcement, plus external red teamers. On training data, the card names three categories: select publicly available data, proprietary data from partnerships, and human data from trainers, red teamers and employees.
Sora 2 and the app
On September 30, 2025, OpenAI announced Sora 2 as its "flagship video and audio generation model," released alongside a new Sora app. OpenAI said it was more physically accurate, realistic and controllable than earlier systems and added synchronized dialogue and sound effects. Its illustration of the physics claim: if a basketball player misses a shot, the ball rebounds off the backboard instead of teleporting into the hoop, as the company said earlier models might do. The page now carries a notice that the Sora product is no longer available as of April 26, 2026.
What the shutdown involved
The help center page lists two dates and a few practical points:
| Item | What OpenAI says |
|---|---|
| Web and app | Discontinued April 26, 2026 |
| API | Discontinued September 24, 2026 |
| Export | Content can be exported at sora.chatgpt.com/sunset by clicking Export |
| Data | Permanently deleted after discontinuation and any final export window |
| Credits | Purchased ChatGPT/Sora credits can still be used for Codex |
The API documentation corroborates the second date. Its deprecations table lists the Videos API, sora-2, sora-2-pro and dated snapshots with a September 24, 2026 shutdown date, a March 24, 2026 notice date and no one-to-one replacement. That is six months between notice and shutdown, which is a useful reference point when you judge any video API.
What to take from Sora's history
If you evaluate a text-to-video tool today, the Sora record suggests a short checklist. Ask for the maximum duration and resolution in writing, since the 20-second, 1080p figures were the product maximum (Plus was limited to 480p and 720p) and the one-minute figure applied to a research model. Check whether outputs carry provenance metadata such as C2PA. Find the export path for your own files before you build a library inside any app. And read the provider's deprecation policy, because a product that was a headline launch in 2024 was gone by the spring of 2026.
Take this further with Oakgen Agent
Agent can run web searches, open vendor pricing and docs pages, and write a sourced comparison table or a downloadable spreadsheet.
Looking for a recap of one specific YouTube video? The YouTube Summarizer does that today.
Open this task in Oakgen Agent
The prompt is filled in for you. Nothing is sent until you press send.
Frequently asked questions
Is OpenAI Sora still available?
No. According to OpenAI's help center, the Sora web and app experiences were discontinued on April 26, 2026, and the Sora API was discontinued on September 24, 2026. OpenAI's API deprecations page lists the Videos API and the sora-2 models with a September 24, 2026 shutdown date and no one-to-one replacement. This was checked on 2026-11-19.
How long and how sharp could Sora videos be?
At the December 2024 launch OpenAI said users could generate videos up to 1080p resolution and up to 20 seconds long, in widescreen, vertical or square aspect ratios. Those were the headline maximums, not the Plus tier: the same announcement says ChatGPT Plus included up to 50 videos at 480p, or fewer videos at 720p, each month, and that Pro added higher resolutions and longer durations. The February 2024 technical report said the largest research model could generate up to a minute of high-definition video.
What did Sora 2 add?
OpenAI's September 30, 2025 announcement described Sora 2 as a video and audio generation model with more physical accuracy and controllability than earlier systems, including synchronized dialogue and sound effects. It was released together with the Sora app.
Can I still export videos I made in Sora?
OpenAI's help center says content can be exported from sora.chatgpt.com/sunset by clicking Export, after which an email arrives when the export is ready. It also says that after discontinuation, and after any final export window, data associated with Sora use will be permanently deleted. Check the help center for the current status of the export window.
Sources
- OpenAI Help Center, What to know about the Sora discontinuation (read 2026-11-19)
- OpenAI, Sora is here, December 9, 2024 (read 2026-11-19)
- OpenAI, Sora System Card, December 9, 2024 (read 2026-11-19)
- OpenAI, Video generation models as world simulators, February 15, 2024 (read 2026-11-19)
- OpenAI, Sora 2 is here, September 30, 2025 (read 2026-11-19)
- OpenAI API documentation, Deprecations (read 2026-11-19)


