The Videos API goes away on 24 September and OpenAI names nothing to move to.

OpenAI notified developers on 24 March 2026 that the Videos API, along with the sora-2 and sora-2-pro model aliases and their snapshots, would be removed from the API on 24 September 2026. The deprecation entry lists no successor model or endpoint.

That absence is the whole story. Look down the rest of the deprecations page and every entry pairs a shutdown with a migration: the reasoning models point at the GPT-5.6 family, the image models point at gpt-image-2, the transcription models point at gpt-transcribe. This one points nowhere.

A deprecation with a named successor is usually a string change plus a regression test. You swap the model identifier, re-run your evaluations, check that output shape and latency are close enough, and ship. The work is bounded and you can estimate it.

A deprecation with no successor is a vendor selection exercise. You have to survey what else exists, test output quality against your actual prompts rather than a demo, compare pricing models that may not be shaped the same way, check terms for your use case, and then do the integration. That is a different order of work, and it cannot be compressed into the week before the deadline.

As of publication that leaves seventeen days. If video generation is a feature your users can reach, this is the item on the calendar to deal with first, ahead of things with later dates but easier migrations.

The endpoint and the model names are both greppable, and the surface area is usually small because video generation tends to be one feature rather than a cross-cutting dependency.