Nobody is leaving Sora. It left. OpenAI closed the app on 26 April 2026 and the API on 24 September, and no Sora model survives inside ChatGPT. What replaces it depends on which part you used: the cameo feed, the realistic clips with sound, or the API.
also searched as: sora alternative · sora alternatives · alternative to sora · apps like sora ai · free sora alternative · alternatives to sora ai · sora ai alternatives free
Stay on Sora if you only need your old videos back, because the one thing still running at sora.com is an Export button, and OpenAI says it permanently deletes your Sora data once any final export window has passed.
checked 2026-10-13 · Sora is Discontinued by OpenAI: the Sora app and website closed on 26 Apr 2026 and the Sora 2 API on 24 Sep 2026 (OpenAI Help Center). There is no plan to buy and no Sora model in ChatGPT; only a data export remains. · by dan scutaru
better: The nearest thing to Sora's feed with a cameo button. Remixing a Short with Google's Omni model is free in the YouTube app and YouTube Create, and "Add me to this scene" drops your avatar into eligible Shorts, where the audience already is.
you give up: Avatars need you to be 18, own a channel and live in the US, Canada, Australia or New Zealand. Only you can use your avatar, so friends starring in each other's clips is gone. Avatar clips carry a visible watermark, and creators can opt out of being remixed.
better: Its "Your likeness" setting is the closest copy of Sora's cameo permissions: only you, approved followers, mutual follows or everyone can put you in a video, and you are told when someone posts one. Free, with Vibes as the feed and its own app.
you give up: A video starts as an image you animate, with music or a voiceover added, not a prompt that writes its own dialogue. Free limits are unpublished; more costs $7.99 a month on Meta One Core. Depending on your region, your AI use shapes the ads Meta shows you.
better: The Gemini app now makes video with Google's Omni model, which Google says replaces Veo there: 10-second clips with sound, up to five photos as input, edits in plain words, and your own avatar via @ and your Google username. From $4.99 a month on Google AI Plus.
you give up: No video on the free plan, and no published daily clip count on any tier. 18 and over only. Editing a video you upload is blocked in the EEA, Switzerland, the UK and some US states.
better: Kling 3.0 makes 3 to 15 seconds with native audio and spoken dialogue in five languages, plans multi-shot sequences itself, and locks a character's face and voice across shots from 2 to 4 reference images. Standard lists at $10 a month.
you give up: Free output is 720p, watermarked and not licensed for commercial use, and the free monthly credit grant has no published size. Sound costs half as much again: 9 credits a second at 720p against 6 without.
better: ByteDance's model takes up to 12 references at once, mixing photos, video and audio clips, cuts multi-scene videos with its own transitions, and generates synced dialogue and effects. Dreamina hands out free daily credits to try it.
you give up: It blocks real faces in reference images and video, so you cannot star in it the way you did in Sora. Free downloads carry a Dreamina watermark, and the final price only appears after you sign in.
better: The most like-for-like swap. Veo 3.1 Fast is $0.10 a second at 720p with audio, what sora-2 cost, Lite is $0.05, and extensions chain a clip to 148 seconds. Google now recommends Omni Flash as the default, at about $0.10 a second.
you give up: Veo 3.1 is still labelled preview, and preview models can be pulled. Clips are 4, 6 or 8 seconds, there is no free tier, and finished videos are deleted from Google's servers after 2 days.
better: One API key for several vendors' models, so the next shutdown is a config change rather than a rewrite. It lists Veo 3.1 at Google's own $0.40 a second with audio, Kling 3 Pro at $0.14 a second, and Seedance 2.0.
you give up: Nothing for non-developers: no app, no editor. No free credits on its pricing page, and units differ by model, Seedance being billed per thousand tokens, so comparing costs takes arithmetic.
better: Its API sells Runway's own Gen-4.5 at $0.12 a second next to Veo 3.1, Seedance 2.0, Gemini Omni Flash and Grok Imagine, all on one $0.01 credit. A separate web app, from $12 a month billed yearly, gives the rest of the team an editor.
you give up: Kling is in Runway's app but not on its API price list. Veo 3.1 costs the same $0.40 a second with audio as going to Google. Free web accounts get 125 credits once, never refilled.
better: Apache 2.0 open weights from Alibaba. The 5B model makes 720p at 24fps on a single 24GB card such as an RTX 4090, with no credits, no account and nothing leaving your machine.
you give up: Slow: Alibaba quotes up to 9 minutes for a 5-second 720p clip on one consumer GPU, and the 14B models want 80GB. No generated soundtrack. The open line stops here: no Wan 3.0 weights are published.
better: Open weights that generate picture and sound together, the thing Sora 2 was known for, plus audio-to-video, regenerating one section of a clip, and 4K output. Free for anyone under $10 million in annual revenue.
you give up: The current LTX-2.5 download is about 66GB and wants a large Nvidia card; FP8 and offloading help smaller ones, slowly. The weights sit behind accepting Lightricks' licence on Hugging Face, which is not an open-source licence.
OpenAI told developers on 24 March 2026 that Sora was ending. The app and website closed on 26 April, and the Sora 2 models and Videos API went dark on 24 September. OpenAI's model page says there is no one-to-one replacement, and its Sora 2 launch post now opens by saying the product is no longer available.
What is left is an export page. On 13 October, sora.com redirects to a screen that says Sora is no longer available, with one Export button. Nothing replaced it inside ChatGPT: the release notes through 9 October list no video generation.
If Sora was a feed you scrolled and a cameo you lent to friends, the nearest apps are Meta AI, home of the Vibes feed and likeness permissions, and YouTube, where Shorts Remix runs Google's Omni model at no cost. Both are on iPhone and Android.
For better clips rather than a feed, Gemini on Google AI Plus and Kling both have phone apps. Grok Imagine is no longer the cheap option: since 12 October, video in Grok needs X's $30-a-month Premium plan.
Three need no card: Shorts Remix in YouTube, Meta AI, and Kling's monthly free credits. Each charges you another way. YouTube watermarks every remix and links it back to the original, Meta does not publish its limits, and Kling's free clips are 720p, watermarked and not for commercial use. Dreamina's free Seedance credits also leave a watermark.
Two that look free are not. Gemini's free plan makes no video, and Runway's 125 free credits arrive once and never refill. The only free with no strings on the output is Wan 2.2 or LTX-2 on your own graphics card, which costs you the card and the wait.
sora-2 and sora-2-pro have been unavailable since 24 September, and OpenAI names no successor. sora-2 cost $0.10 a second at 720p. The same money buys Veo 3.1 Fast with audio, or about the same for Gemini Omni Flash, and Veo 3.1 Lite halves it. Runway's Gen-4.5 is $0.12, and Kling 3 Pro is $0.14 on fal.
Whatever you pick, put a layer between your code and the model. Sora gave six months' notice; the next shutdown may give less. A router such as fal or Runway's API turns it into a one-line model swap.
No, because there is nothing left to use, free or paid. The app, free tier included, closed on 26 April 2026 and the API on 24 September. OpenAI says credits bought for ChatGPT or Sora can still be spent on Codex.
OpenAI's spokesperson pointed to focus and growing compute demand, and said the Sora research team would keep working on world simulation for robotics. Reporting at the time added falling downloads and Disney dropping a planned $1 billion investment.
Yes, for now. sora.com opens an export page, and OpenAI emails you when your file is ready. Its help center says all Sora data is permanently deleted after any final export window, so export now rather than waiting for that email.
No. Sora was OpenAI's only video product, and ChatGPT's release notes through 9 October 2026 list no video generation. ChatGPT can write the script and the prompts; the frames have to come from another tool on this page.