Sora 2 Explained: Price, Access, and Why It Was Shut Down

Sora 2 was OpenAI's video-and-audio generation model and the invite-only social app built around it, launched on September 30, 2025 — but if you're searching for it today, the most important thing to know is that it's gone. OpenAI's own product pages now carry a standing notice: "As of April 26, 2026, the Sora product is no longer available." The Sora app and sora.com went dark on that date, and the developer-facing Sora API followed on September 24, 2026. Here's what Sora 2 actually was, what it cost while it existed, how the invite and cameo system worked, how it differed from the original Sora, and where to look now that it's discontinued.

What Was Sora 2?

OpenAI described Sora 2 as its "flagship video and audio generation model" — a successor to the original Sora that could follow more intricate instructions across multiple shots, model physical interactions more accurately (a missed basketball shot rebounds off the backboard instead of teleporting into the hoop, in OpenAI's own example), and generate synchronized audio, including ambient sound, dialogue and sound effects, alongside the video itself. OpenAI framed the leap from Sora 1 to Sora 2 as roughly analogous to the jump from GPT-1 to GPT-3.5 in language models: a step change in what the system could reliably do, not just a quality bump.

Sora 2 wasn't released as a bare model. OpenAI paired it with a new, standalone social app simply called "Sora," where people could generate clips, remix other users' generations, browse a personalized feed, and — most distinctively — drop themselves or consenting friends into AI-generated scenes.

Under the hood, both Sora generations built on the same basic idea OpenAI laid out in its original February 2024 technical report: treat video the way large language models treat text, breaking clips into "spacetime patches" that a diffusion transformer learns to generate, rather than training on fixed-size clips. That original research version of Sora could already produce up to a minute of video and images up to 2048x2048, with emergent behaviors like object permanence and rough 3D consistency appearing purely from scaling up training compute — years before Sora 2 added synchronized audio and consent-based likeness on top of that foundation.

Is Sora 2 Still Available in 2026?

No. This is the answer most people searching "Sora 2" in late 2026 actually need. OpenAI's own Sora 2 announcement page, its Sora 2 system card, and its safety write-ups ("Launching Sora responsibly" and "Creating with Sora safely") all carry the identical footnote: the Sora product stopped being available on April 26, 2026, roughly seven months after launch. The Sora iOS and Android apps were pulled, and sora.com stopped serving the product. Consistent with that, the app's former App Store listing now returns a "page doesn't exist" error rather than a working download link.

Developer access lasted a little longer. OpenAI's platform documentation lists the Sora video-generation models and the Videos API itself under its deprecations table, with a shutdown date of September 24, 2026, and no recommended replacement model listed. OpenAI had notified API users of the planned removal back on March 24, 2026 — the same day reporting emerged that OpenAI was ending the consumer product — giving developers roughly six months' notice before the Videos API endpoints, and the sora-2 and sora-2-pro model aliases, stopped working entirely.

OpenAI has not published a dedicated blog post walking through its full reasoning for the shutdown. Its public materials note only that the Sora research team would continue focusing on world-simulation research aimed at robotics, rather than continuing to operate the consumer app and API.

Cameo and Characters: How Sora 2's Signature Feature Worked

The feature people associated most with Sora 2 was the ability to put a real person into a generated video. OpenAI initially called this "cameo" at launch and later renamed it "characters." The mechanic, per OpenAI's own descriptions, worked like this: a user recorded a short one-time video-and-audio clip inside the app to verify their identity and capture their likeness and voice. Once that character was created, only that person decided who else could use it — they could revoke access, delete any video that included their character (even ones drafted by other users), and review every video that used their likeness at any time.

OpenAI layered extra safeguards specifically onto character-based videos: stricter moderation thresholds than ordinary generations, blocks on depicting public figures outside of the consented-character system, visible and dynamically-moving watermarks on shared outputs, and C2PA provenance metadata embedded in every generated video so it could be traced back to Sora. Teen accounts had further limits, including a non-personalized feed option and parental controls managed through ChatGPT.

Sora 2 Price and Access While It Was Live

At launch, Sora 2 was free to use, with what OpenAI called "generous limits" that were still ultimately bounded by available compute. Access rolled out through invitations, starting in the U.S. and Canada, with OpenAI stating an intent to expand to more countries over time; the app later reached additional markets including Japan before it was discontinued.

A separate, higher-quality model — Sora 2 Pro — was reserved for ChatGPT Pro subscribers and was made available first on sora.com, with OpenAI saying it planned to bring Sora 2 Pro into the Sora app as well. OpenAI's own post did not publish a standalone monthly price for a "Sora" subscription; Sora 2 Pro access rode on top of the existing ChatGPT Pro plan. For current ChatGPT plan pricing and what each tier includes today, see this site's comparison of every ChatGPT, Claude and Gemini plan and price, since subscription pricing across OpenAI's product line has continued to change since Sora 2's shutdown.

OpenAI also said its only concrete monetization plan at the time was a possible future option to pay for extra generations if demand outstripped available compute — a plan that, per the product's own after-the-fact status notice, never got the chance to mature into a lasting business.

The Sora 2 API: What Developers Could Do

OpenAI said at launch that it planned to release Sora 2 in the API, separate from the consumer app, and it eventually did — offering the sora-2 and sora-2-pro model aliases (OpenAI's deprecations table also lists dated snapshots, including sora-2-2025-10-06 and sora-2-2025-12-08, among the models removed). The Videos API let developers generate video programmatically rather than through the Sora app's interface, which is what powered a wave of third-party tools and integrations built on top of Sora 2 during its roughly year-long developer availability window.

That access ended on September 24, 2026. OpenAI's deprecations documentation lists the Videos API's shutdown with no successor model in the "recommended replacement" column, meaning developers who had integrations built on Sora 2 had to migrate to a different provider entirely rather than a newer OpenAI video model. If you're evaluating OpenAI's broader API roadmap and pricing after Sora's exit, this site's breakdown of GPT-6's API prices and ChatGPT access covers what OpenAI has continued to actively support.

Sora 1 vs. Sora 2: What Actually Changed

Sora's history runs through two distinct public releases before the shutdown. The original Sora was previewed as research in February 2024 and reached general availability as "Sora Turbo" on December 9, 2024, folded into ChatGPT Plus and Pro at no extra cost rather than shipping as its own app. Sora 2 was a different kind of release: a new model plus a dedicated, invite-based social app.

OpenAI's Sora 2 launch post noted that "Sora 1 Turbo will remain available" alongside the new model at the time — both products were ultimately wound down together when the whole Sora line was discontinued in 2026.

Content Policy and Safety Notes

Because Sora 2 could place real people's faces and voices into generated video, OpenAI's safety documentation focused heavily on consent and provenance. Every generated video carried visible and invisible (C2PA metadata) provenance signals; shared videos with a person's likeness carried dynamically moving watermarks including the creator's name. Uploads of photorealistic people, and any video involving a minor or young-looking person, were subject to stricter moderation than ordinary text-to-video prompts. OpenAI also said its systems automatically scanned generated speech for policy violations and blocked attempts to generate music imitating living artists or existing copyrighted work, and that it would honor takedown requests from creators who believed a Sora output infringed on their material.

None of this is retrievable through the product today — the policies above describe how Sora 2 operated while it was live, not a system you can currently test.

Sora 2 Alternatives: What's Still Active

With Sora 2 gone and no OpenAI successor named, people searching for a working AI video generator in late 2026 are largely looking outside OpenAI's product line. Google's Veo model family, accessible through the Gemini app and API, and Runway's Gen-series models are among the actively maintained text-to-video and image-to-video tools still being developed by their respective companies. Neither is an OpenAI product, so this site does not carry verified pricing or feature specifics for them here — check each company's own site for current plans. For a sense of how OpenAI's still-active multimodal tools compare on the image-generation side, see this site's explainer on Google's Nano Banana Pro, one of the image models frequently compared against OpenAI's own image tools.

What to Do Now

If you're looking for Sora 2 today: it isn't there. Downloading "Sora" from either app store won't work, since the listing itself no longer resolves, and sora.com no longer serves the product. If you built anything on the Sora API, OpenAI's own deprecations table confirms the September 24, 2026 cutoff already passed with no replacement model named, so any integration depending on sora-2 or sora-2-pro needs a new video-generation provider. If you're mainly interested in what OpenAI still actively supports, its ChatGPT and GPT-6 API lines remain live, and this site's other explainers on current OpenAI pricing and access are the more useful next stop than searching for a way back into a discontinued product.

Written byTheo Park

Theo Park runs the AI desk at Pandromeda. He follows model launches from the frontier labs and the open-weight community, tracks the assistants and developer tools built on them, and explains what each release changes on pricing, capability and safety. His reporting leans on primary sources: model cards, technical reports, API documentation and the companies' own announcements.

More from Theo ParkAll AI coverage