Sora Review 2026: Discontinued, What Happened and What to Use Instead

What it is

Sora was OpenAI's text to video model. The first version was previewed in February 2024 and opened to ChatGPT Plus and Pro subscribers in the US and Canada on December 9, 2024.

Sora 2 arrived on September 30, 2025 with a new standalone iOS app, followed by Android about two months later. The app worked like a short video social network: an algorithmic feed with what OpenAI called steerable ranking, and cameos, a feature that let you put yourself or consenting friends into generated scenes. Sora 2 generated video with synced audio from text or an image, and developers could call it through the Videos API.

Seven months after the Sora 2 launch, the app was gone. This page keeps the full record for anyone researching what happened, migrating an API integration, or comparing today's tools against what Sora offered.

Sora timeline

Date	Event
February 15, 2024	Sora previewed with demo videos up to one minute long
December 9, 2024	Sora opens to ChatGPT Plus and Pro users in the US and Canada
September 30, 2025	Sora 2 and the invite only iOS app launch
October 6 to 7, 2025	MPA chairman criticizes the opt out copyright approach
October 7, 2025	404 Media reports watermark removal sites
October 30, 2025	Japan's CODA asks OpenAI to stop using members' works
December 11, 2025	Disney announces a $1 billion investment and character license
March 24, 2026	OpenAI announces Sora's discontinuation
April 26, 2026	Sora app shuts down
September 24, 2026	Videos API, sora-2 and sora-2-pro shut down

Pricing explained

There is nothing to buy in October 2026. The prices below are the last ones OpenAI published, kept for comparison and for teams estimating what a replacement will cost.

The consumer app launched free, with what NBC News described as generous limits, and OpenAI said it might later let users pay for extra videos during high demand. The API was metered per second of output.

Model (API)	Resolution	Price per second	8 second clip	20 second clip
sora-2	720x1280 or 1280x720	$0.10	$0.80	$2.00
sora-2-pro	720x1280 or 1280x720	$0.30	$2.40	$6.00
sora-2-pro	1024x1792 or 1792x1024	$0.50	$4.00	$10.00
sora-2-pro	1080x1920 or 1920x1080	$0.70	$5.60	$14.00

How that compares now. Google's Veo 3.1 Fast costs $0.12 per second at 1080p through the Gemini API, so an 8 second 1080p clip is $0.96, against $5.60 on sora-2-pro. Veo 3.1 Standard at $0.40 per second is still cheaper than sora-2-pro at 1080p. Veo 3.1 Lite at $0.05 per second for 720p undercuts the old sora-2 rate by half. If cost drove your Sora usage, you will likely pay less elsewhere. See the Veo review and the AI Price Index.

Rate limits before shutdown. The sora-2 model page listed 50 requests per minute on the Build tier, 200 on Launch and 375 on Grow.

Key features

These are the features Sora 2 shipped with, as documented by OpenAI's model pages and launch coverage.

Synced audio

Sora 2 generated dialogue, effects and ambient sound together with the picture. That was its headline advantage at launch. Veo 3.1 and Kling now offer native audio too, so this is no longer a reason to wait for Sora.

Cameos and likeness consent

To create a cameo, you recorded yourself on video and audio once. OpenAI's stated rule, reported by NBC News, was that a likeness could become a cameo only with that person's consent, and the consent could be revoked at any time. It was the most consent focused likeness system of the mainstream video apps, and its loss leaves a gap: Google Flow avatars cover only your own likeness and are not available in the EEA, UK or Switzerland.

Social feed

The Sora app was built around a feed of generated clips with steerable ranking, so you could tune what it showed you, much like TikTok. This made it a social network first and a production tool second, which matters for why it failed: casual viewing did not pay for the compute.

Resolutions and API

The API supported 720p on sora-2 and up to 1920x1080 on sora-2-pro, with text or image input, via the v1/videos endpoint only. The last snapshot was sora-2-2025-12-08.

Licensed characters (announced, never shipped at scale)

Disney's December 2025 agreement was to bring 200+ Disney, Pixar, Marvel and Star Wars characters to Sora and ChatGPT images starting in early 2026, with curated Sora videos streaming on Disney+. The agreement excluded talent likenesses and voices.

Commercial rights and licensing

Sora's short life was dominated by rights disputes. Here is what the record shows.

Opt out training and generation. Sora 2 launched letting users generate copyrighted characters unless rights holders opted out. On October 3, 2025, OpenAI said it would give rights holders more granular control. Days later, the chairman of the Motion Picture Association criticized the opt out approach and said OpenAI bears responsibility for preventing infringement, per Deadline's October 7, 2025 report as cited on Wikipedia.

Japan's CODA request. On October 30, 2025, the Content Overseas Distribution Association, whose members include Studio Ghibli, Square Enix, Bandai Namco and Sony's Aniplex, asked OpenAI to stop using members' works without permission for Sora 2 training and generation.

Disney deal and guild response. Disney announced a three year license and a $1 billion equity investment on December 11, 2025. SAG-AFTRA, the Writers Guild of America and The Animation Guild objected in a joint letter, and Variety reported the WGA's view that the deal appeared to sanction what it called theft of its members' work. After the shutdown announcement, TechCrunch reported Disney learned of it less than an hour beforehand and the partnership was terminated.

Lawsuits. We found no court case naming Sora specifically in the sources we read. The disputes above were letters, public statements and contract changes, not filed suits.

Provenance. Every Sora 2 video carried a visible moving watermark and C2PA metadata identifying it as AI generated. 404 Media reported on October 7, 2025 that at least three websites removed the watermark in seconds. If provenance matters to you, pick a tool with an invisible watermark that survives editing, such as Google's SynthID on Veo, and do not rely on visible marks alone.

Ownership of old outputs. Videos you made are your files. Keep local copies; with the app closed, the hosted versions are gone.

What users say

There are no live ratings to cite. The official Sora app is no longer listed on the US App Store, and searching "Sora" now returns look alike apps from unrelated developers, such as "SorAI" with a 3.8 average from 1,220 ratings. Be careful: several websites and apps use the Sora 2 name to sell access to other models.

The usage numbers tell the story. TechCrunch reported that Sora peaked at about one million users and fell below 500,000 before the shutdown, while costing about $1 million a day to run. Early praise centered on realism, physics and synced audio; criticism centered on deepfake risk, copyright and a feed full of low effort clips.

Privacy and security

With the app closed, the practical question is your data. OpenAI's help center published a discontinuation notice in March 2026; its pages block automated access, so we could not read the export and deletion terms directly. Check your OpenAI account's data controls for anything still stored, and assume cameo recordings were tied to your OpenAI account.

For any replacement, compare training defaults. Google's Gemini Apps use activity for training while Keep Activity is on, and the paid Gemini API does not use prompts for training. Kling's terms let it train on your content unless you email to revoke.

Who it is for

• Researchers and journalists documenting the Sora story, the copyright fights and the Disney deal.
• Developers with dead Sora 2 API calls who need the old price and resolution specs to scope a migration.
• Teams evaluating AI video that want to know what Sora offered before picking a replacement.

Who should skip it

Everyone who wants to make video. Sora is not available in the app, in ChatGPT or through the API. Also skip any site selling "Sora 2" access in October 2026; OpenAI shut the model down, so those sites are reselling something else or nothing at all.

Alternatives

• Veo: the closest replacement, with native audio, 4K, SynthID and API prices from $0.05 per second. Read the Veo review.
• Kling: longer clips up to 15 seconds per generation and plans from $10 per month, with strict branding and commercial use terms. See the Kling review.
• Runway: an editing focused studio with more control tools. See the Runway review.
• Luma Dream Machine and Pika for quick, cheaper experiments.
• CapCut if you mostly need editing with some AI generation built in.

The full ranked list is on Sora alternatives and best AI video generators.

Verdict

Sora 2 showed what consumer AI video could look like: synced audio, consent based cameos and a social feed of generated clips. It also showed the economics do not work for a free social app burning about $1 million a day. OpenAI closed the app on April 26, 2026 and the API on September 24, 2026, and did not name a successor. Move to Veo 3.1 if you want the closest quality and the clearest provenance, Kling if you want longer clips for less, and Runway if you want editing control.