OpenAI discontinued the Sora app on April 26, 2026 and shut down the Sora API on September 24, 2026, without naming a replacement. As of October 1, 2026, the closest match for most Sora users is Gemini Omni Flash in Google Flow: text-to-video with native audio, an avatar feature like Sora's characters, 50 free credits a day and paid plans from $4.99 a month. For image-to-video, use Kling 3.0 or MiniMax H3; developers can move to the Gemini, Kling, MiniMax, Runway or xAI APIs, all billed per second of video.
Checked on October 1, 2026 against OpenAI's help center and API docs and each alternative's official pricing, plan and API pages. Prices are in US dollars and change often; the linked pages are the source of truth.
Sora shutdown 2026: what happened and when
OpenAI posted "We're saying goodbye to Sora" on X on March 24, 2026, then closed the product in two stages without naming a replacement. The Wall Street Journal reported that OpenAI was winding down products that use its video models and would not offer video generation inside ChatGPT either.
| Date | What happened |
|---|---|
| March 24, 2026 | OpenAI announced the shutdown and notified developers that the Videos API and all Sora 2 models were deprecated. |
| April 26, 2026 | The Sora web and app experiences were discontinued. |
| September 24, 2026 | The Videos API and the sora-2, sora-2-pro and dated snapshot models were shut down. OpenAI lists no recommended replacement. |
| After any final export window | OpenAI says it will permanently delete all data associated with your use of Sora. |
Sources: OpenAI's Sora discontinuation article, its API deprecations page and the Sora 2 launch post, which now reads "As of April 26, 2026, the Sora product is no longer available." An OpenAI spokesperson told Engadget that the Sora research team would focus on world simulation research for robotics.
What people used Sora for, and what replaces each job
Sora 2 launched on September 30, 2025 with synchronized dialogue and sound effects, inside a social app built around remixes and a feature called characters that dropped you into any scene after a one-time recording. No single tool replaces all of it, so match the replacement to the job:
| If you used Sora for... | Switch to | Why |
|---|---|---|
| Text-to-video clips with sound | Gemini Omni Flash or Veo 3.1 in Google Flow | Native audio, 50 free credits a day, paid plans from $4.99 |
| Putting yourself in scenes (characters) | Google Flow avatars | Record your likeness once, then type @me in a prompt |
| Animating your own images | Kling 3.0 or MiniMax H3 | First and last frames, reference characters, clips up to 15 seconds |
| Remixing other people's videos | YouTube Shorts Remix with Gemini Omni | Free; remix an eligible Short with your own prompts and images |
| Editing your own clips | Gemini Omni Flash | Edits a generated or uploaded video through follow-up prompts |
| The Videos API in an app | Gemini, Kling, MiniMax, Runway or xAI APIs | Asynchronous jobs billed per second, like Sora |
The best Sora alternatives in October 2026
| Tool | Best for | Clip length | Audio | Entry price | Commercial use |
|---|---|---|---|---|---|
| Google Flow (Gemini Omni Flash, Veo 3.1) | Text-to-video, avatars, editing | Omni Flash 4 to 10 s; Veo 3.1 4 to 8 s | Yes | Free: 50 credits a day. Google AI Plus: $4.99 a month for 200 more | Google says it won't claim ownership of what you generate |
| Kling 3.0 | Image-to-video, recurring characters | 3 to 15 s | Optional | $6.99 first month, then $8.80 a month for 660 credits | Paid plans yes; free Basic plan no |
| MiniMax H3 (Hailuo) | Image-to-video and reference-to-video, 2K | 4 to 15 s | Yes | API: $0.08 a second at 768p | Paid Hailuo plans yes; free downloads are watermarked |
| Runway Gen-4.5 | Editing-heavy projects | 2 to 10 s at 720p | No | $15 a month for 625 credits (about 52 s of Gen-4.5) | No non-commercial restriction |
| Wan 2.2 (local) | Free, private, unlimited runs | About 5 s per run | No | Free on your own GPU | Apache 2.0 |
Sources: Flow plans, Kling membership, MiniMax pricing, Runway pricing, Runway commercial use and Wan 2.2. Leaderboard positions are in our AI video generator ranking.
Text-to-video with sound: Gemini Omni Flash and Veo 3.1 in Google Flow
Google Flow is the closest consumer replacement for the Sora app. Every account gets 50 credits a day, which do not roll over; Google AI Plus ($4.99 a month) adds 200 a month and AI Pro ($19.99) adds 1,000. A Gemini Omni Flash clip at 720p costs 7, 10, 12 or 15 credits for 4, 6, 8 or 10 seconds, and a Veo 3.1 Lite clip costs 10 (credit costs). Without a subscription, you may not be able to generate video between roughly 2 PM and 5 PM UTC.
- Avatars replace Sora's characters. Flow lets you create an avatar with your own visual and audio likeness on a phone, then add it to any prompt by typing @me. Avatars are not available in the EEA, the UK or Switzerland, and avatar data can't be shared through Flow Tools or public project links.
- Ask for one shot if you need one. Google's Omni Flash prompt guide says the model tries to create several shots by default, so write "in a single continuous shot" when you want one take.
- Watermarks and rights. Google's Flow FAQ says every Veo and Omni output carries an invisible SynthID watermark, the visible watermark is a toggle except in India, South Korea and Vietnam, and Google won't claim ownership of what you generate.
Image-to-video: Kling 3.0 and MiniMax H3
Kling VIDEO 3.0 makes 3- to 15-second clips with optional native audio, takes start and end frames, and keeps a recurring character consistent with Elements built from 2 to 4 reference images or a short video. It costs 6 credits a second at 720p without audio and 12 at 1080p with audio. Kling 4.0, due in October, raises the limit to 30 seconds; our Kling 4.0 guide tracks what is live, and the Kling tutorial covers prompts and settings.
MiniMax H3 handles text-to-video, image-to-video from a first or last frame, and reference-to-video, at 768p or 2K for 4 to 15 seconds, and MiniMax says it generates native stereo sound. In the Hailuo app, paid plans download without a watermark, and Hailuo's terms say it does not claim ownership of what paid subscribers generate, including the right to commercial use. Our image-to-video guide compares more tools.
Sora alternatives that are free
- Google Flow's free plan: 50 credits a day buys three 10-second Omni Flash clips at 720p or five Veo 3.1 Lite clips, and the plan list includes avatars at no cost. 1080p upscaling starts with Google AI Plus.
- YouTube Shorts: YouTube says Remix with Gemini Omni is free in Shorts and the YouTube Create app, so you can remix an eligible Short with your own prompts and images, such as inserting yourself alongside a creator. The YouTube app also makes standalone clips up to 10 seconds from a prompt, except in the European Economic Area.
- Wan 2.2 on your own GPU: the open-weight models are Apache 2.0, their developers claim no rights over your output, and they generate at 480p or 720p. The repository's reference script for the 5B model needs a 24 GB GPU. Our ComfyUI guide explains the usual setup.
- Kling's free Basic plan: fine for tests, but the membership page reserves commercial use, 1080p and watermark removal for paid plans.
Daily limits, watermarks and commercial terms for more free tools are compared in our free AI video generators guide.
Sora API alternatives: endpoints and prices per second
The Sora API billed $0.10 a second for sora-2 at 720p and $0.30 to $0.70 a second for sora-2-pro, depending on resolution. Most replacements work the same way: you create a job, then collect the file.
| API and model | Create a video | Get the result | Length | Price per second |
|---|---|---|---|---|
| OpenAI Sora 2 (shut down) | POST /v1/videos | GET /v1/videos/{video_id}/content | 4 to 20 s | sora-2 $0.10; sora-2-pro $0.30 (720p) to $0.70 (1080p) |
| Gemini API: Gemini Omni 1.1 Flash | POST /v1beta/interactions | Returned in the response (base64 or a URI) | Up to 10 s in Flow; extensions to 40 s | About $0.10 at 720p, with audio |
| Gemini API: Veo 3.1 | POST /v1beta/models/veo-3.1-generate-preview:predictLongRunning | Poll the operation, download within 2 days | 4, 6 or 8 s; extensions to 148 s | Standard $0.40; Fast $0.10; Lite $0.05 (720p, with audio) |
| Kling 3.0 | POST /text-to-video/kling-3.0 or /image-to-video/kling-3.0 | GET /tasks or a callback URL; files cleared after 30 days | 3 to 15 s | $0.084 without audio, $0.126 with audio (720p) |
| MiniMax H3 | POST /v2/video_generation | GET /v2/query/video_generation/{task_id} within 7 days | 4 to 15 s | $0.08 at 768p, $0.13 at 2K |
| Runway | POST /v1/text_to_video or /v1/image_to_video | GET /v1/tasks/{id} | Gen-4.5: 2 to 10 s | Gen-4.5 $0.12, silent; Gemini Omni Flash $0.10 |
| xAI Grok Imagine Video 1.5 | POST /v1/videos/generations | GET /v1/videos/{request_id}; temporary URLs | 1 to 15 s | $0.14 at 720p ($0.08 at 480p, $0.25 at 1080p) |
Sources: OpenAI video guide, Omni Flash and Veo docs, Gemini API pricing, Kling API and pricing, MiniMax API, Runway API and pricing, and xAI video docs and pricing.
What that means per clip: an 8-second 720p video cost $0.80 on sora-2. The same clip costs about $0.81 on Gemini Omni Flash, $0.80 on Veo 3.1 Fast, $0.40 on Veo 3.1 Lite, $0.64 on MiniMax H3 at 768p, $1.01 on Kling 3.0 with audio and $1.12 on Grok Imagine Video 1.5. At 1080p, 8 seconds cost $5.60 on sora-2-pro, against $3.20 on Veo 3.1 Standard and $1.34 on Kling 3.0 with audio. These are our calculations from the published rates, before failed attempts.
Sora API migration checklist
- Recover what you can. The download route,
GET /v1/videos/{video_id}/content, went with the Videos API on September 24; OpenAI's policy is that a shut-down endpoint "will no longer be accessible." Sora's download URLs lasted only an hour, so look in the storage your pipeline copied files to, and keep the prompts, sizes and durations you logged. Content made in the Sora app goes through the export at sora.chatgpt.com/sunset. - Map each feature you used. A first-frame
input_referencebecomes Kling'sfirst_frame, a MiniMaximage_urlwith thefirst_framerole, or an image input to Veo or Omni Flash. Reusable characters map to Kling Elements (up to 3 per request), Veo reference images (up to 3) or MiniMax reference-to-video. Extensions map to Veo 3.1 (7-second steps up to 148 seconds) or Omni 1.1 Flash (10-second steps up to 40), and edits to Omni Flash follow-up turns withprevious_interaction_id. - Fit the durations. Sora took 4 to 20 seconds; Veo 3.1 takes 4, 6 or 8, Kling 3.0 3 to 15, MiniMax H3 4 to 15 and Grok Imagine 1 to 15. Split longer shots and join them in an editor.
- Replace webhooks. Kling and MiniMax accept a
callback_url; Veo, Runway and xAI return an operation or ID to poll; Omni Flash can return the video in the same response. - Re-price and save outputs. Kling's API needs a prepaid package (from $700 for 5,000 units, valid 180 days), and Runway sells API credits at $0.01 each. Download every result at once: Veo deletes files after 2 days, MiniMax tasks can be queried for 7 days, Kling clears results after 30 days, and xAI URLs are temporary.
- Recheck content and data rules. Sora's API rejected real people and input images with human faces. Veo allows only adults in image-to-video, with stricter person settings in the EU, UK, Switzerland and MENA, and Omni Flash does not accept uploads of certain recognizable people. On the Gemini API, free-tier content is used to improve Google's products and paid-tier content is not, and the API terms say Google won't claim ownership of generated content.
- Retest your prompts. Omni Flash composes several shots unless told otherwise, and Kling's API turns on
multi_shotby default. Run your most common Sora prompts on two candidates before moving production traffic.
Build a Video Pipeline That Survives the Next Shutdown
Sora's closure showed the cost of building on one model. Keeping your character in reusable stills and reference packs means you can move to whichever video model leads next. Our AI Influencers course covers that production pipeline, from character design in ComfyUI to image-to-video and platform disclosure rules.
AI Influencers: The Full Production Pipeline
Character consistency, image-to-video workflows with hosted and open models, publishing and the disclosure rules that apply to AI content.
Get AI Influencers →Sora Shutdown and Alternatives FAQ
When did the Sora app shut down?
OpenAI discontinued the Sora web and app experiences on April 26, 2026, less than seven months after the Sora app launched on September 30, 2025. The Sora API and the sora-2 and sora-2-pro models followed on September 24, 2026.
Is Sora AI shut down for good?
Yes. OpenAI's API deprecations page lists the Videos API and every Sora 2 model as shut down on September 24, 2026, with no recommended replacement. OpenAI's policy is that a shut-down model or endpoint is no longer accessible.
What is the best Sora alternative?
For most former Sora users, Gemini Omni Flash in Google Flow: text-to-video with native audio, an avatar feature similar to Sora's characters, 50 free credits a day and a $4.99 Google AI Plus plan, as of October 1, 2026. To animate your own images, use Kling 3.0 or MiniMax H3. Through the Gemini API, Omni Flash costs about $0.10 a second at 720p, the same rate as sora-2.
Are there free Sora alternatives?
Yes. Google Flow gives every account 50 credits a day, enough for three 10-second Gemini Omni Flash clips at 720p. YouTube says Remix with Gemini Omni is free in YouTube Shorts and the YouTube Create app, and the open-weight Wan 2.2 models run free on your own GPU under the Apache 2.0 license.
Can I still download my Sora videos?
For content made in the Sora app, OpenAI's help center points to sora.chatgpt.com/sunset to request an export and says Sora data will be permanently deleted after any final export window. Videos made through the API could only be downloaded through the Videos API, which shut down on September 24, 2026, so only files you already saved remain.
Why did OpenAI shut down Sora?
OpenAI said: "As we focus and compute demand grows, the Sora research team continues to focus on world simulation research to advance robotics." The Wall Street Journal reported that the shutdown was part of a refocus on business and coding tools ahead of a possible IPO.