Sora API Shuts Down September 24: Export First — OpenAI Lists No Replacement

Sora API Shuts Down September 24: Export First — OpenAI Lists No Replacement

Export the keepers. Do not wait for a Sora successor. OpenAI’s Help Center says the Sora web and app experiences ended on 26 April 2026, and the Sora API shuts down on 24 September 2026. The API deprecations table lists the Videos API, sora-2, sora-2-pro, and the named snapshots with no recommended replacement. ClipCanva is not affiliated with OpenAI, Google, Canva, Runway, or Kling. If a campaign still lives only as a Sora job id, pull the files from sora.chatgpt.com/sunset, lock the claim on the AI script generator, then rebuild clip-length beats on the AI video generator or image to video. After 24 September, a landing page that still says “powered by Sora 2” is advertising a dead endpoint.

Shutdown facts (confirmed 15 September 2026)

Source	Public claim	Workflow implication
OpenAI Help: Sora discontinuation	Web and app experiences discontinued 26 April 2026. Sora API discontinued 24 September 2026. Export at sora.chatgpt.com/sunset. Purchased ChatGPT/Sora credits can still be used for Codex.	Credits are not a video runway. They do not keep /v1/videos alive.
Same Help article	After shutdown, and after any final export window, OpenAI will permanently delete data associated with Sora use.	Treat unexported clips as gone. Do not assume a second window.
OpenAI API deprecations	Notice dated 24 March 2026. Removal on 2026-09-24 for Videos API, sora-2, sora-2-pro, sora-2-2025-10-06, sora-2-2025-12-08, sora-2-pro-2025-10-06. Recommended replacement column is ---.	There is no official OpenAI video successor to migrate the same call to. Rebuild the packet on another desk.
OpenAI Videos API guide	Same shutdown list. While the API is still up: sora-2 for speed, sora-2-pro for higher quality and 1080p (1920x1080 or 1080x1920). Both support 16- and 20-second generations. Jobs are async (POST /videos, poll, then download). Guardrails reject copyrighted characters/music, real people, and input images with human faces.	If you still render in the next nine days, use it to export missing keepers — not to start a new always-on pipeline.
Canva Help: Create a video clip	Canva AI Create a video clip: 8 seconds with audio, 6 seconds without. English prompts only. Optional reference image. Wait up to 2 minutes. Separate from Magic Media text-to-video and Magic Video assembly.	A clip desk is short. Split a 16–20s Sora beat into two or three cards.
Google Flow creative controls	27 August 2026: Gemini Omni 1.1 Flash in Flow adds start and end frames, 1080p/4K export, and faster lower-credit 360p drafts.	First/last frames replace “hope the next Sora extension matches.” Draft cheap, then keep.
Kling 3.0 on Runway	Text or image in; optional first and last frames; native audio and speech on Pro; Standard / Pro / Turbo tiers.	A live third-party desk with start/end control. Confirm the live tier before you quote audio.

Confirm the live generator before you quote seconds, resolution, or credit cost. OpenAI’s replacement cell is empty on purpose. Do not invent a secret Sora 3.

What dies on 24 September — and what does not

Surface	Status after 24 September 2026	Operator move
sora.com / Sora apps	Already gone since 26 April 2026	Use the sunset export, not the old app
Videos API (POST /videos and related)	Scheduled removal	Stop new production jobs; drain the queue
sora-2, sora-2-pro, listed snapshots	Scheduled removal, no replacement listed	Copy prompts and first frames out of the vendor chat
Finished MP4s you already downloaded	Yours, if you have the files	Relabel by shot, not by job id
ChatGPT/Sora credits	Help says they can still be used for Codex	Do not budget them as video seconds
A wrapper site still labeled “Sora 2”	Not an OpenAI route after shutdown	Treat it as a different model or as false advertising

Sora 2 was a job API: create, optionally guide with an image, extend, edit, download. It was not a timeline editor and not a script desk. If your “workflow” was one prompt and a webhook, you do not have a packet. You have a dependency.

Pick the desk before you rebuild

The fail is treating “find a Sora alternative” as the same job as “ship the same ad.” Alternatives lists answer which model. This week answers which file you still have and which desk that file belongs on.

Job	Where it actually runs	What you must supply	What the system may invent	Fail condition
Legal line, claim, spoken words	AI script generator	Audience, one claim, duration, forbidden words	Extra scenes, unearned superlatives	Prompting video from a Sora logline you never signed
Shot list / clip-length beats	Script output + a scene card you keep	Shot size, camera move, cut, seconds	Decorative adjectives that fight the still	One 20s Sora paragraph dumped into an 8s clip desk
Still that must stay on-model	Stills desk, then image to video	Last good Sora frame or a new approved photo as first frame	Camera path, warped type, extra logos	Animating a compressed social download of the wrong SKU
Finished campaign clip	AI video generator or a named model on ClipCanva Compare	Locked still and/or locked caption, crop, seconds	Extra props, a second spoken line	Shipping take one because it “feels like Sora”
Cheap motion test	Flow 360p draft (documented Google path) or a low-res batch pass	Same first frame and caption as the keeper	Texture, lighting drift	Spending 4K credits on a caption you have not locked
Existing Sora or plate footage you must match	AI video summarizer then script	Source cut, what must stay from the plate	New blocking the plate does not support	Asking a new model to “just continue the Sora clip” with no card
Reusable camera language	Prompt ideas	Shot type, lighting, camera, duration	A chat you will lose on 24 September	Changing five variables on every retry
Short social assembly from clips you already have	Canva Magic Video (assembly, not a generator)	Clips or photos you already approved	Template, music, a longer cut	Treating assembly as text-to-video

If the brief is “this bottle already exists; make it move 8 seconds,” that is image-to-video from a locked still. If the brief is “we do not know the shots yet,” that is script. If the brief is “keep the Sora character, change only the eyeline,” that is one-variable control, and most clip desks will fail it unless you pin a first frame. Name the desk before you open a tab.

Sora API vs clip desk vs first/last-frame desk vs assembly

Dimension	Sora Videos API (until 24 Sep)	Clip desk (Canva-class)	First/last-frame desk (Flow / Kling-class)	Assembly (Magic Video-class)
Official surface	OpenAI /v1/videos	Canva AI Create a video clip	Google Flow; Kling 3.0 on Runway; ClipCanva generators	Canva Magic Video and similar
Input	Prompt; optional image; async job	English prompt; optional still; wait	Prompt plus pinned frames	Clips or photos you already have
Documented length	16s and 20s supported on Sora 2 / Pro while live	8s with audio / 6s silent on Canva’s clip desk	Confirm live; do not copy Sora’s 20s onto another button	About 60s sequenced cuts on Canva Magic Video
Audio	Generated with the clip while the API is up	Optional; you cannot manually change in-clip audio on Canva’s clip desk	Native audio/speech on Kling 3.0 Pro (per Runway)	Music/templates on files you fed it
Human gate	Guardrails: no real people, no copyrighted music/characters, faces on input images rejected	Identity, label, extra limbs	Identity across the two stills	Continuity of the files you fed it
After 24 Sep	Dead	Still a clip desk	Still a control desk	Still assembly
Do not use it for	A permanent production dependency	Writing the plot	Inventing the pack shot	Inventing a product that was never shot

Canva’s own split is a useful third map, not a partnership. Create a video clip waits, then hands you 6 or 8 seconds. Magic Media text-to-video sits inside a design. Magic Video sequences footage you already have. Flow’s start and end frames are the documented Google version of “where does the camera need to go?” Kling 3.0 on Runway is the documented third-party version of the same pin. ClipCanva’s image to video is the operator version: a first frame you chose, plus a caption that came from a shot, not from a mood board.

Nine-day workflow: export, packet, then batch

Write the packet once. Generate on a desk that will still exist on 25 September.

1. Export before you brainstorm. Open sora.chatgpt.com/sunset, click Export, wait for the email. Download MP4s and stills onto disk you control. Help’s language is blunt: export as soon as possible; deletion follows shutdown and any final window.
2. Inventory by shot, not by vibe. For each keeper: duration, crop, first frame, last frame, spoken line, what must not change (label, hands, logo). If you only have a compressed social post, it is a reference, not a master.
3. Rewrite the blueprint on a script desk. On the AI script generator, lock audience, one claim, close, duration, and forbidden words. A Sora prompt that already worked is still a prompt. It is not a signed line.
4. Split to the next desk’s cap. Canva’s clip desk is 8 seconds with audio or 6 without. Sora’s 16–20s beat is two or three cards, not one overloaded caption. Park reusable camera language in prompt ideas.
5. Lock a still before motion. Pull a clean frame from the export, or shoot/generate a new hero. Send that still to image to video. If there is no still, generate on the AI video generator only after the caption is a shot.
6. Use first and last frames when the path matters. Flow documents start/end frames and a 360p draft before 1080p/4K. Kling 3.0 on Runway documents pinning first and last frames. Do not spend the keeper credit on a caption you have not locked.
7. Compare models when the first desk warps the label. Use ClipCanva Compare. Do not start a new story in the second model. “Feels like Sora” is not an acceptance test. Readable type is.
8. If you have plate or old Sora footage, run it through the AI video summarizer so the new script respects what was actually on screen. Do not ask a new model to extend a file you cannot re-render.

Operator card you can paste:

Deadline: Sora API 24 Sep 2026 — no official replacement
Export: sora.chatgpt.com/sunset (files on disk, not a job id)
Output: 8s 9:16 Reel, English, one claim
Desk: script → shot card → still → batch image-to-video
Blueprint: [one sentence story]
Shot: MCU, slow push, label readable last frame
First frame: sora_export_frame_03.png (legal signed)
Caption: [shot card, not “make it like Sora”]
Forbidden: extra logos, competitor names, medical claims, real-people likeness
Fail if: set restyles; label unreadable; new VO; wrapper still named Sora after 24 Sep
Retry: same first frame, change one variable
Not this job: new production on /v1/videos; one-box cinematic prompt with no shot list

Why the empty replacement cell is the story

OpenAI did not point sora-2 at a next alias. The deprecations page leaves the cell blank. That is the production fact.

A prompt that says “cinematic, highly detailed, masterpiece” does not tell the next vendor where the camera goes. A 20-second Sora take will not paste into Canva’s 8-second clip desk. A still you never saved will look expensive in memory and still be the wrong SKU.

Guardrails on the dying API are also not a style. No real people, no copyrighted music, no faces on input images — those were OpenAI’s rules. The next desk has its own. Re-check likeness and music before you regenerate.

Credits moving to Codex is the other half. Help is explicit: leftover ChatGPT/Sora credits can be used for Codex. They are not a bank of video seconds you can spend after 24 September.

Creator checklist

• [ ] You exported from sora.chatgpt.com/sunset and have MP4s on disk you control.
• [ ] Each keeper has a shot card: size, move, duration, first frame, spoken line.
• [ ] You wrote a script with one claim before you opened a replacement video model.
• [ ] Duration fits the live desk (Canva clip: 8s with audio / 6s silent — confirm yours). A 16–20s Sora beat is split, not squeezed.
• [ ] The first frame is an approved still, not a thumbnail from a social re-upload.
• [ ] The caption is the shot card, stored in prompt ideas if you will reuse it.
• [ ] You did not budget ChatGPT/Sora credits as future video renders.
• [ ] One variable changes per retry. “Make it like Sora” is not a variable.
• [ ] Morph, extra limbs, new on-frame type, or a second spoken line is a fail.
• [ ] Plate or old Sora footage was read with the AI video summarizer before you asked a model to continue it.
• [ ] You did not leave production on /v1/videos after 24 September, and you did not trust a wrapper that still brands itself Sora 2.

FAQ

When does Sora actually stop?

Two dates. The web and app experiences stopped on 26 April 2026. The Sora API stops on 24 September 2026. That includes the Videos API and the sora-2 / sora-2-pro aliases and snapshots listed on OpenAI’s deprecations page. ClipCanva is not part of that shutdown.

Is there an official OpenAI replacement?

Not on the public deprecations table. The recommended replacement column is empty (---). Help says leftover credits can go to Codex. Codex is not a video generator. Plan as if there is no drop-in /v1/videos successor.

Should I keep generating on Sora until the 24th?

Only to fill holes in the export — a missing master, a cleaner frame, a take you never downloaded. Do not start a new always-on pipeline on a dated endpoint. Rebuild keepers on the AI video generator or image to video with a shot card that will survive the cutover.

How long should each rebuilt clip be?

Depends on the desk. Canva Help documents 8 seconds with audio or 6 seconds without for Create a video clip. Sora’s own guide still documents 16- and 20-second generations while the API is live. That length does not transfer. Split the blueprint to the cap you can actually export on 25 September.

When do I compare models?

When the shot card is stable and the first desk warps identity, type, or the product. Compare on ClipCanva Compare. Do not use a second model to rewrite the story, and do not use “Sora-like” as the brief.

The API has a date. The files do not, if you export them. The script is the blueprint. The shot is the prompt. The still is the approval gate. Rebuild on a desk that will still answer on 25 September.