Sora shuts down in 3 days: your migration plan

Updated May 11, 2026 — historical record: Sora’s web and mobile apps went offline on April 26, 2026, as scheduled. The Sora API remains operational until September 24, 2026. Export endpoints are now closed for the consumer product. The original article below (published April 23) and the April 25 follow-up describe the lead-up to the shutdown and the migration plan that was active at the time — both retained as historical record.

The April 23 article below described the situation as of three days before shutdown.

Export your stuff now

Go to sora.chatgpt.com/exports/me and request your data export. The export includes your generated clips, drafts, and account activity. Do this today rather than tomorrow — server load is going to spike as the deadline approaches, and OpenAI’s export queues have been known to take 24-48 hours.

After April 26, the web and mobile experiences are gone. The API still works until September, but without the web app you can’t access existing saved work through the UI.

Why OpenAI is killing a product Disney invested $1 billion in

The numbers tell the short version. OpenAI reports Sora’s user base halved from roughly 1 million to under 500,000 active users in under a year. Compute costs stayed high. Copyright litigation mounted — in particular, the ability to generate videos featuring copyrighted characters became an ongoing legal and brand-safety headache.

The Disney deal is the painful part. In December 2025, Disney announced a $1 billion investment in OpenAI that would let Sora 2 users generate “more than 200 copyrighted Disney characters” — Pixar, Marvel, Star Wars, the Disney Animation catalog. Four months later, Disney reportedly walked back the deal and Sora is being sunsetted. The two events aren’t unrelated.

OpenAI’s stated reason is a “strategic redirect toward enterprise and productivity tools ahead of a potential IPO.” The real story is probably some combination of: usage wasn’t there to justify the compute spend, Disney got cold feet, and Runway / Veo 3 / Kling consolidated the “AI video for professionals” market while OpenAI was focused elsewhere.

What to use instead

Three serious options in April 2026:

Runway — the professional AI video leader. Gen-4.5 produces cinematic output. $12 Standard plan is extremely credit-limited (~25 seconds of Gen-4.5 video/month). Pro at $28 is where most creators should land. If you were making ads or short films in Sora, Runway is the closest replacement.

Veo 3 (Google) — the only mainstream AI video model that generates video with synchronized native audio (dialogue, music, environmental sound) in a single pass. Free tier gives 10 generations/month per Google account. Google AI Pro at $19.99 bundles Veo 3.1 with the rest of Gemini. Best fit if your Sora work involved voiceover or music.

Kling — the affordable long-duration champion. Up to 2-minute single-take clips (4x longer than Runway or Veo). $6.99/month Standard plan is the cheapest credible paid tier in the category. Kling 2.5 Turbo matches competitors on quality; 2.6 adds native audio. Chinese-origin caveats apply for sensitive commercial work.

Quick decision tree:

• Cinematic quality matters most → Runway Pro ($28/mo)
• Need voiceover/music inside the generation → Veo 3 via Google AI Pro ($19.99/mo)
• Budget and long clips matter more than polish → Kling Standard ($6.99/mo)
• You only need a few clips a month casually → Veo 3 free tier or Kling free tier

Full ranked comparison at best AI video tools.

What won’t work

Don’t bet on Sora 3. OpenAI has not announced a successor. Given the IPO framing, I’d assume the video generation bet is dead at OpenAI for the next 12-18 months at minimum — compute constrained, not strategically prioritized.

Don’t try to extend the Sora API past September 24. The API deprecation has a hard date. If you’ve built a product on top of the Sora API, you have five months to migrate. Runway, Kling, and Veo all have production APIs that can serve as replacements.

The broader lesson

The AI category is going to have a lot more of these sunsets. Tools that launched in the “pre-consolidation” era of 2023-2024 will keep getting discontinued as vendors focus on the products that actually have enterprise traction. Sora was the flashiest launch of 2024. Eighteen months later it’s a cautionary tale.

If you’re an individual user or a small team, the lesson is: don’t build critical workflows on a single generative-AI product unless that product is in the vendor’s “main line.” Sora wasn’t. Runway, Veo, and Kling are their makers’ main bets — much safer foundation for a workflow you depend on.

If you’re a vendor: shipping a $1B Disney deal doesn’t save a product users aren’t using. The market kills things, not executives. Sora was killed by its own user numbers, not by Sam Altman.

Back up your clips today. Pick a replacement. Move on.

Questions or corrections? Email Pick Right. Want the full list? See all news.