Sora 2 Pricing After Shutdown: Costs and Migration

Sora 2 no longer has an active OpenAI API price to buy. OpenAI's video-generation documentation confirms that the Sora 2 models and Videos API shut down on September 24, 2026. It also says there is no one-to-one replacement API. The web and app experiences ended earlier, on April 26, according to the Sora discontinuation notice. Those are two different retirement dates.

This guide was corrected on October 22, 2026. Its previous headline advertised a pre-shutdown per-second price. That is no longer a usable purchase decision, so the old rate table and instructions to start new Sora jobs have been removed. The useful task now is to preserve your production records, understand what your previous costs mean, and move a shot brief to a supported workflow without assuming equivalent settings or results.

If you arrived looking for a free Sora account, a new Sora API integration, or an upgrade to a higher Sora resolution, this page cannot provide one. A third-party listing with a familiar model name is not proof that OpenAI's retired endpoint is available.

Read old Sora prices as historical records

A price screenshot needs four pieces of context: provider, date, model version, and billing unit. A consumer subscription, an API charge per generated second, and a reseller's credit package are different purchases. A number copied from one surface should not be used to estimate another.

For past work, reconcile the actual invoice or usage export with your generation log. Record the duration billed for each attempt, whether a batch rate applied, and whether the result was accepted. Include discarded variations in production cost. Do not multiply an old headline rate by your final edited video's runtime and call that the whole project cost.

For future work, start a new budget using the replacement service's current terms. Check its charging unit, output size, duration choices, audio support, and failed-job policy before a trial. Save the terms you actually relied on with the project. A cheap first generation can become an expensive workflow if it repeatedly changes the product or needs extensive repair.

Preserve the parts of the project you can reuse

Check the export instructions in OpenAI's discontinuation notice for your account. Do not assume a permanent export window or promise a client that unavailable assets can be recovered. Keep your own downloaded originals, final edits, approved reference images, scripts, captions, and rights records together. A prompt alone is not a production archive.

Create one row per shot in the blank shot migration inventory. It separates the old provider settings from the visual requirement. That distinction prevents an obsolete parameter from becoming a hidden requirement in the next tool.

• Keep: the approved subject, action, composition, delivery ratio, reference files, and review notes.
• Recheck: model names, duration menus, image inputs, audio behavior, download access, and cost.
• Retest: motion, object identity, text, faces, sound timing, and the transition into the next shot.

For an API integration, also inventory queued jobs, retry rules, callbacks, stored output URLs, and downstream editing steps. Stop sending new requests to a retired service. A replacement should be a deliberate workflow change with explicit failure handling, not a silent model-name substitution.

Worked example: move a product shot, not a model setting

Imagine a six-second clip of an unbranded ceramic cup on a kitchen counter. This is a planning example, not a measured model comparison. The approved requirement is a stable cup shape, a readable handle, a gentle camera push, and enough clean space for a caption added in an editor.

The old request might combine a model identifier, a resolution flag, a duration parameter, and a long scene description. Keep the scene requirement separate. The replacement's interface decides which technical settings are valid; the prompt describes only what the viewer should see or hear.

If the selected workflow supports a reference image, use the approved product photo. If it cannot accept the input you need, it is a poor match for this shot regardless of its headline quality. Do not compensate by adding unsupported command-line flags to the prose.

Review the first, middle, and last frames, then watch the complete clip. A correct opening frame does not excuse a changing handle halfway through. If the camera move damages the object, simplify the motion before increasing resolution or duration. If product text matters, keep it out of a moving close-up until the workflow has demonstrated that it can preserve it.

Compare cost per accepted shot

Use a fixed brief and a limited test budget. Record every attempt, including failures. Decide the acceptance checklist before looking at the outputs so an attractive result does not quietly lower the standard. Compare the complete usable asset, including any separate audio or repair work it requires.

For example, if a hypothetical test spends $6 across all attempts and produces three approved clips, generation spend per accepted clip is $2. That arithmetic is not a Sora or QuestStudio price quote. It also excludes editing labor and any other paid tools. If no clip passes, report zero accepted clips and the actual spend; dividing by zero does not produce a useful success metric.

Keep generation cost and subscription allowance estimates separate. The generation cost calculator can help organize a scenario, but your provider's usage records establish actual charges. Repeating the same exercise on the same brief is more informative than comparing unrelated showcase videos.

How QuestStudio helps with the next workflow

Open Video Lab to review the models and controls currently exposed for your account. Select a supported image-to-video or text-to-video route that matches the brief, review its usage requirements, and test one short shot. This link is a workflow starting point, not a promise that Sora is available or that another model is a drop-in replacement.

Keep approved prompts and revisions organized in Prompt Lab. For the final edit, use your timeline editor to assemble cuts, captions, and audio. Preserve the original generated file so export compression can be diagnosed separately from generation defects.

Frequently asked questions