Sora video creation is AI-generated video: a person describes a scene in words, and OpenAI’s Sora model generates a video from that prompt. Some Sora 2 documentation also describes image or video inputs. Availability depends on the channel: OpenAI says its Sora web and app experiences ended on April 26, 2026, and its Sora API ended on September 24, 2026. (OpenAI Help Center)
What does Sora video creation do?
Sora turns a written description of a scene into moving images. OpenAI introduced the original Sora as a text-to-video model. Demonstrations included scenes such as people walking through snowy Tokyo; examples illustrate a reported output, not what every prompt will reliably produce. (OpenAI)
OpenAI described Sora 2 as a video-and-audio generation model. Microsoft’s Azure/Microsoft Foundry documentation separately described Sora 2 workflows for text, image, and video inputs, audio output, and Remix for targeted edits. These are channel-specific documented capabilities, not a guarantee that every Sora interface supported them. (OpenAI; Microsoft Learn)
What are Sora’s limitations?
Generated video is not guaranteed to match every detail in a prompt. Microsoft’s Sora 2 documentation warns about difficulty with complex physics, causal relationships, spatial reasoning, and precise event sequencing. Treat generated clips as drafts to review when movement, object placement, or continuity matters. (Microsoft Learn)
Microsoft’s Sora 2 API documentation describes an asynchronous workflow—submit a generation job, check its status, then retrieve the result—and labels the experience as a preview. Its page gives differing API examples and technical specifications, so those details should not be treated as one universal, current API contract. (Microsoft Learn)
Can developers use Sora for a production API integration now?
Not through OpenAI’s Sora API: OpenAI’s Help Center states that the API was discontinued on September 24, 2026. It gives April 26, 2026 as the end date for the Sora web and app experiences. (OpenAI Help Center)
Azure/Microsoft Foundry is a separate, less-settled case. Microsoft’s current catalog lists “Sora” (sora-2) as a model, while noting that model availability varies by region and cloud. However, Microsoft’s Sora 2 video-generation guide labels the API preview, and the public catalog listing alone does not confirm that a particular version can currently be deployed in a given subscription—or that it is production-ready. Microsoft’s retirement schedule does not list Sora in the retrieved table; Microsoft Q&A also contains a November 7, 2026 user report of deployment errors for two Sora 2 versions. Taken together, public information does not establish a reliable, currently deployable Azure version or its retirement conditions. Do not treat Azure Sora 2 as a confirmed production option without checking the exact version and deployment eligibility in Microsoft’s current service documentation and your subscription. (Microsoft Foundry model catalog; Microsoft Learn retirement schedule; Microsoft Q&A)
The Azure video-generation guide says billing is per second but gives no numeric rate or cost example, so it is not enough to estimate spend. (Microsoft Learn)
Bottom line
Sora video creation means generating video from a prompt; some Sora 2 documentation also describes reference-media inputs, audio, and targeted edits. OpenAI’s Sora API has ended. Microsoft still lists sora-2 in its catalog, but that listing does not settle version-level deployability or production readiness. Confirm those conditions before choosing Azure Sora 2 for a new integration.
Sources
What to know about the Sora discontinuation — OpenAI Help Center
Sora: Creating video from text — OpenAI
Sora 2 video generation overview (preview) — Microsoft Learn
Foundry Models sold by Azure — Microsoft Learn
Model retirement schedule — Microsoft Learn
Sora-2 deployment blocks on ServiceModelDeprecated error — Microsoft Q&A