Does OpenAI have a video generator?

OpenAI’s consumer video story is short and specific: Sora launched on September 30, 2025, the consumer app shut down on April 26, 2026, and the Sora API closes on September 24, 2026. OpenAI does not offer a consumer video generator as of September 2026. A successor model called Spud has been pre-trained but is being positioned for industrial and robotic physical-world-understanding applications, not consumer video.

Did OpenAI ever have a video generator?

OpenAI’s text-to-video model was called Sora. First previewed on February 15, 2024, Sora launched as a limited-access consumer app in the US and Canada on September 30, 2025. A second version, Sora 2, followed within months, adding synchronized audio and a TikTok-style social feed with a “cameos” feature. The app let users generate video clips of up to 20 seconds from a text prompt.

For most of its consumer lifetime (September 2025 through April 2026), Sora was OpenAI’s official answer to the “how do we generate video?” question. The company stopped offering that answer in April 2026.

When did OpenAI’s Sora shut down?

The shutdown happened in stages through 2026.

Content created inside the Sora app had to be downloaded before the April 26 cutoff. Any content held in Sora API accounts is scheduled for permanent deletion after the September 24 API shutdown.

Why did OpenAI shut down Sora?

Four connected reasons, based on OpenAI’s own shutdown announcement and reporting that followed.

Compute economics: Sora reportedly cost OpenAI roughly $15 million per day in inference compute at peak. Total lifetime subscription revenue: approximately $2.1 million. That is a two- to three-orders-of-magnitude gap between running cost and revenue.

Engagement drop: Sora’s monthly active user count dropped from roughly one million at launch to under 500,000 by early 2026. Users found the 20-second cap and content restrictions too limiting for sustained daily use.

Copyright pressure: Sora was in negotiations for a $150 million video-generation deal with Disney. Disney killed the deal in early 2026 citing unresolved rights concerns around the model’s training data.

Strategic refocus: OpenAI is preparing for a public offering and is redirecting compute toward enterprise and productivity tools (coding assistants, ChatGPT enterprise features) that have durable paid revenue. Consumer video did not fit the near-term revenue thesis.

What is Spud, OpenAI’s successor?

Spud is the internal name for OpenAI’s next video model. Sam Altman has publicly confirmed that pre-training is complete. The model exists.

The important caveat: Spud is not being positioned as a consumer TikTok-style video app. OpenAI has framed it for “physical world understanding” applications: warehouse robotics, driverless-vehicle simulation, industrial process modeling. Not creator tools, not marketing video, not YouTube content.

OpenAI has not announced a consumer-facing successor to Sora, and Spud’s messaging so far explicitly rules out that positioning. If you were waiting for a Sora 3 to fill the consumer video gap, current signals say that is not what is coming next from OpenAI.

What AI video tools should you use instead of Sora?

Four short-form generative models currently produce the video Sora was best at, plus one category Sora never targeted.

Short-form generative models (Sora replacements for social clips):

• Google Veo 3: the most direct Sora replacement. 8-second clips at up to 1080p, native audio, strong motion fidelity. Available through Google AI Studio and Gemini.
• Kling 3.0: 5 to 12 second clips at 1080p, native audio, strong on cinematic realism.
• MiniMax Hailuo: the audio-driven video workhorse, particularly strong on lip sync and character animation.
• Runway Gen-4: established option for creators, well-integrated with editing workflows.

Long-form explainer platforms (a category Sora never targeted):

• Scenema: 5 to 20 minute finished video from a text prompt, with recurring characters, single-pass narration, and multi-scene structure. See best AI tools for long-form explainer videos for a full ranked comparison.

If you were using Sora for short social clips, Veo 3 or Kling is the closest replacement. If you needed long-form output (which Sora never delivered well), Scenema or another long-form platform is a stronger fit today than Sora ever was.

Can you still access Sora?

Not through the consumer app or the web experience. The Sora app has been dark since April 26, 2026. The Sora API is still functional at the time of this article’s publish date (September 23, 2026) but shuts down permanently on September 24, 2026. After that date, no OpenAI infrastructure will serve Sora inference. Any content held in API accounts is scheduled for permanent deletion.

The TL;DR on OpenAI’s video generator

Does OpenAI have a video generator? Not as of September 2026. Sora, its previous consumer video model, was shut down in stages through 2026. A successor model (Spud) is pre-trained but is being positioned for industrial applications, not consumer video.

When did Sora shut down? The consumer app on April 26, 2026. The API on September 24, 2026. Shutdown announcement was March 24, 2026.

Why did OpenAI shut down Sora? Four connected reasons: roughly $15M/day compute costs against approximately $2.1M lifetime revenue, engagement drop from one million to under 500,000 users, a killed $150M Disney deal over rights concerns, and strategic refocus on enterprise revenue ahead of a public offering.

What is Spud? OpenAI’s next video model. Pre-training is complete. Positioned for physical-world-understanding applications (industrial, robotic, simulation) rather than consumer video.

What should I use instead of Sora? For short clips: Google Veo 3 or Kling 3.0. For long-form explainer video: Scenema. For talking heads: HeyGen or Synthesia.

Where to go next

• Read can ChatGPT make videos for the related disbelief-cluster question about ChatGPT’s own video capabilities.
• See best AI tools for long-form explainer videos for a ranked comparison of what currently produces AI video.
• Read how to make a 6-minute AI explainer video from a short prompt for a walkthrough of Scenema’s long-form pipeline.
• See what is a long-form explainer video for the definition of the format Scenema targets.
• Try Scenema at scenema.ai.