Short answer: OpenAI has fully shut Sora down. The Sora app and website closed on April 26, 2026, and the Sora API followed on September 24, 2026. The closest replacement for most people is Google Veo 3.1, which generates video and synced audio in one pass. Runway Gen-4.5 is the better pick if you edit footage as much as you generate it, and Seedance 2.5 is worth a look if you need longer single shots. You can use all three, plus Luma, Hailuo, Vidu, Pika and Higgsfield, from one LazySusan account.
If you're here because a Sora integration quietly started failing two weeks ago, the rest of this post is for you.
What actually happened to Sora
OpenAI announced the end of Sora on March 24, 2026, and shut it down in two stages:
- April 26, 2026: the Sora app and sora.chatgpt.com went dark.
- September 24, 2026: the Sora API was switched off.
OpenAI asked people to download their videos before the cutoffs. If you didn't, I wouldn't count on getting them back. OpenAI said it would email users if a later export window opened, so check the inbox tied to your account before you give up.
On LazySusan, Sora is gone from the video menu too. It isn't coming back unless OpenAI ships a new video model, and there's no sign of that yet.
Which Sora alternative should you use?
Here's the quick version. I've grouped these by the job you're trying to do, because "best AI video generator" isn't a useful question on its own.
| If you want to... | Use | Why |
|---|---|---|
| Make a short clip with sound from a text prompt | Veo 3.1 | Video and synced audio in one generation, good prompt following |
| Draft lots of ideas cheaply | Veo 3.1 Fast or Veo 3.1 Lite | Same family, a fraction of the cost per second |
| Edit footage you already shot | Runway Aleph 2 | Swap objects, relight, restyle an existing clip |
| Generate cinematic shots with fine camera control | Runway Gen-4.5 | Strong motion and physics, text or first-frame image input |
| Get one long continuous shot | Seedance 2.5 | Reported to generate up to 30 seconds in a single pass |
| Animate a still image of a person | Hailuo 2.3 | Good body movement and facial expression |
| Get stylized or social-first clips fast | Pika, Vidu, Higgsfield | Quick, cheap, fun effects and templates |
Veo 3.1: the default replacement
If you used Sora mainly for "type a prompt, get a clip with sound," Veo 3.1 is where I'd start. It's the most Sora-like experience: one prompt in, a 4, 6 or 8-second clip out, with dialogue, ambience and sound effects generated alongside the picture.
There are three tiers, and the difference matters for your wallet:
- Veo 3.1 is the full-quality model. Use it for the final version.
- Veo 3.1 Fast is cheaper and quicker. Google cut its price in April, and for most social clips I honestly can't tell the difference on a phone screen.
- Veo 3.1 Lite is the budget tier at 720p or 1080p. It's great for testing a prompt before you spend real money on it.
A workflow that works well: write the prompt, run it three or four times on Lite or Fast, pick the best composition, then do one final render on full Veo 3.1. That's how I'd have used Sora too if it had a cheap tier.
Prompts that work well with Veo are specific about the camera and the sound. Something like:
"Handheld close-up of a barista pouring latte art in a busy cafe, morning light through the window, shallow depth of field. Sound: milk steamer hiss, low chatter, a cup set down on a saucer."
Leave out the sound line and you'll still get audio, but it'll be generic.
If you want more Veo prompt ideas, I wrote a separate post on ways to use Veo for video. It was written for an earlier version, but the prompting advice still holds.
Runway Gen-4.5 and Aleph 2: if you edit, not just generate
Sora was a generator. Runway is closer to an editing suite with generation built in, and for a lot of people that's actually what they needed.
Gen-4.5 makes clips from text or from a first frame you supply. It does well on motion and physical realism, and it led the Artificial Analysis text-to-video leaderboard after launch. If your Sora prompts were about camera moves ("slow dolly in," "drone pull back over the cliff"), Gen-4.5 handles that kind of direction well.
Aleph 2 works on footage you already have. You give it a clip and an instruction like "make it night, add rain" or "replace the red car with a white van," and it changes that while keeping the rest. It handles clips up to about 30 seconds at 1080p. For anyone making product videos or ads from real footage, this is the tool Sora never was.
Seedance 2.5: for longer single shots
ByteDance's Seedance 2.5 came out over the summer. Its headline feature is length: it's reported to generate up to 30 seconds of continuous video in one pass, where most models top out around 8 to 10 seconds. It also generates ambience and sound effects with the picture, and it accepts a lot of reference images, which helps when you need a character or product to stay consistent.
I'd treat the spec sheet with a little caution. Most of the detailed numbers come from launch coverage rather than independent testing. But if you need a long take without stitching clips together, it's the first one I'd try.
Luma Ray, Hailuo 2.3, Vidu, Pika and Higgsfield
These are the "second opinion" models. None of them is a full Sora replacement on its own, but each one is good at something:
- Luma Ray is good for dreamy, cinematic motion and quick image-to-video.
- Hailuo 2.3 from MiniMax is strong at people: body movement and facial expressions. It doesn't generate audio, so you'll add sound separately. ElevenLabs on the same account can handle voice and music.
- Vidu is fast and cheap, good for social content and effects.
- Pika is fun for quick stylized edits.
- Higgsfield focuses on camera-motion presets that look expensive.
The useful thing about having all of them in one place is that you can run the same prompt through three models and keep the best one. Video models disagree with each other a lot more than chatbots do.
What about Kling?
Kling 3.0 is a strong model and a lot of the "Sora alternatives" lists put it near the top, especially on price. I'll be straight with you: Kling is temporarily unavailable on LazySusan right now, so if Kling is the one you want, you'll need to use it directly for the moment. HeyGen, for talking-avatar videos, is in the same temporarily-unavailable bucket.
What does each clip cost?
Video is the most expensive thing you can do with AI, by a wide margin. On LazySusan, you pay with tokens (600 tokens = $1 of model usage), and the video tools show you the token cost before you press generate, so there are no surprises.
To give you a feel for scale, using Google's published per-second rates:
| Model | 8-second clip, rough token cost |
|---|---|
| Veo 3.1 (full quality, with audio) | about 3,800 |
| Veo 3.1 Fast | about 1,000 to 1,200 |
| Veo 3.1 Lite (720p) | about 500 |
Prices for other models vary by length, resolution and whether audio is on, so check the number on screen.
What that means in practice:
- Starter ($8/month, 1,000 tokens) is a text plan. You can try a Lite clip or two, but it won't cover real video work.
- Pro ($19/month, 5,700 tokens) covers a few Lite drafts plus one final Veo 3.1 render, or about a dozen Lite drafts on their own.
- Heavy video work means top-ups. A $25 pack is 15,000 tokens, which is about four full-quality Veo 3.1 clips.
If you were producing lots of Sora videos every week on a flat ChatGPT Pro subscription, I won't pretend tokens will be cheaper for that volume. Pay-per-clip is better if your video use is occasional or bursty, which is true for most people who aren't running a content studio.
How I'd move a Sora workflow over
- List what you used Sora for. Ad clips, b-roll, memes, storyboards. Each one maps to a different model in the table above.
- Rewrite your prompts with sound and camera direction. Sora was forgiving with vague prompts. Veo and Runway reward specifics.
- Draft cheap, finish expensive. Run variations on Veo 3.1 Lite or Fast, then render the keeper on full Veo 3.1 or Gen-4.5.
- Fix audio separately when you need to. If a model doesn't generate audio, or the audio is off, generate a voiceover or music bed with ElevenLabs and put them together in your editor.
- Keep your originals. The Sora shutdown is a reminder that hosted generations can disappear. Download what you make.
You can see every video model LazySusan currently offers on the models page.
FAQ
Is Sora still available anywhere? No. OpenAI closed the Sora app and website on April 26, 2026 and shut off the Sora API on September 24, 2026. Any service still advertising Sora is either out of date or running something else under that name.
What is the closest thing to Sora right now? Google Veo 3.1. It takes a text prompt and returns a short clip with synced audio, which is the experience most Sora users were after. Runway Gen-4.5 is the closest for cinematic camera work.
Can I get my old Sora videos back? Probably not. OpenAI asked users to download their media before the shutdown dates and said it would email users if a later export window opened. If you didn't get an email, assume the files are gone.
Is Kling on LazySusan? Not right now. Kling, HeyGen and PixVerse are temporarily unavailable. Veo 3.1, Runway, Seedance, Luma, Hailuo, Vidu, Pika and Higgsfield are available.
Which Sora alternative is cheapest? Among Google's models, Veo 3.1 Lite is the cheapest per second. On LazySusan an 8-second Lite clip costs roughly 500 tokens, compared with about 3,800 for full-quality Veo 3.1.