OpenAI retires Sora 1 in the U.S. as it consolidates video generation around Sora 2

OpenAI has removed its older Sora 1 video-generation experience in the United States, making Sora 2 the default for U.S. users and ending the option to switch back to the legacy system. The company says Sora 1 was removed on March 13, 2026, and that data export for older content may remain available only for a limited time before permanent deletion.

Sora 1 is gone, and Sora 2 is now the default

The change narrows OpenAI’s consumer video product to a single current experience in the U.S. In the company’s help documentation, OpenAI says Sora now opens in Sora 2 by default and that users can no longer return to Sora 1. The legacy system had supported older video and image generation.

OpenAI also says image generation will no longer be available inside Sora once Sora 1 is removed, though users can still create images in ChatGPT. That leaves Sora 2 as the company’s main path for video generation inside the Sora product.

What users can still recover from the legacy system

For people who created content in Sora 1, OpenAI says exports may still be available for a limited time. The export would include Sora 1, Sora 2 and ChatGPT data, and the company says users can also download individual images and videos from their library while the files remain accessible.

OpenAI says it will email users before the final export window begins, but the company’s guidance is clear: once the deprecation window closes, Sora 1 generations and related social activity will no longer be available through the product.

Why OpenAI is collapsing the product stack

OpenAI says the move is meant to reduce complexity and keep improvements flowing into Sora 2 across web and mobile. In practical terms, that simplifies the product line for both users and the company, but it also shows how quickly frontier generative AI products are moving from experimental, overlapping versions toward a single managed release.

For a video model family still being refined in public view, the decision is operationally important: it limits legacy support, concentrates feedback on one current system and gives OpenAI a cleaner base for further updates to its generative video tooling.