# OpenAI introduced Sora 2 with a separate app and staged invite rollout OpenAI announced Sora 2 on September 30, 2025, pairing a new video-and-audio model with a U.S. and Canada invite rollout, Pro tier and published safety restrictions. Canonical: https://rangoon.ai/insights/archive-openai-sora-2-september-2025/ Author: Rangoon Editorial (https://rangoon.ai/insights/editorial/) Published: 2026-10-24 Updated: 2026-10-24 Event date: 2025-09-30 Tags: [OpenAI](https://rangoon.ai/insights/tags/openai/), [creative AI](https://rangoon.ai/insights/tags/creative-ai/) Source note: Historical reporting: Source 0 is OpenAI’s dated September 30, 2025 Sora 2 announcement. Source 1 is the contemporaneous public system card, used for rollout and safety details. Image: OpenAI introduced Sora 2 with a separate app and staged invite rollout. Original Rangoon explanatory diagram: Source boundary: dated primary release reporting; original editorial analysis. No vendor interface, photograph, measurement result, or endorsement implied. ## Key takeaways - Sora 2 was announced as a video-and-audio model with a separate app rollout. - OpenAI stated an invite rollout in the U.S. and Canada plus an experimental Pro tier. - The system card described restrictions on photorealistic-person image uploads and all video uploads. ## Sora 2 paired a model announcement with a new distribution route [source 1](https://openai.com/index/sora-2/) [source 2](https://openai.com/index/sora-2-system-card/) OpenAI dated its Sora 2 announcement to September 30, 2025 and described a new model for video with synchronized audio. The release was coupled to a new iOS Sora app and an invite-based rollout, initially in the United States and Canada, rather than a statement of immediate global availability. It also mentioned sora.com access after an invite. That distribution detail is central to the event: a model announcement and a social-style product surface arrived together, but access was deliberately staged. The record does not demonstrate that any particular user received an invitation or that clips made in the app were suitable for publication. ## The product statement separated current access from future plans [source 1](https://openai.com/index/sora-2/) [source 2](https://openai.com/index/sora-2-system-card/) OpenAI said Sora 2 would begin with free limits and that Sora 2 Pro was an experimental option for ChatGPT Pro users. It also described an API as a future plan, which is not the same as a released developer endpoint on September 30. These distinctions prevent a common backprojection error: treating the announced model, a Pro experiment and a possible API as one uniformly available service. The announcement supports a specific initial access model with plan and invite boundaries. It does not provide a measured comparison between tiers, a promise that each request would be accepted, or a guarantee that model output would preserve physical, factual or narrative accuracy. ## Safety restrictions shaped the initial feature surface [source 1](https://openai.com/index/sora-2/) [source 2](https://openai.com/index/sora-2-system-card/) The Sora 2 system card says the initial rollout restricted uploads of images containing photorealistic people and restricted all video uploads. It also describes moderation and protections involving minors. These are source-specific conditions, not generic cautionary language: they explain why the September release should not be summarized merely as “upload any media and generate video.” The public safety record gives context for the initial product boundary, while the announcement gives the date and commercial access framing. Neither document supplies a rights clearance for a prompt, image or character. The source evidence supports what OpenAI said it would allow and restrict, not a conclusion about downstream use. ## The news was an invite-stage Sora 2 launch [source 1](https://openai.com/index/sora-2/) [source 2](https://openai.com/index/sora-2-system-card/) Sora 2 should be kept distinct from Sora Turbo, later product changes and current help-page status. The September 2025 event was a named video-and-audio model release with an iOS app, invite rollout, initial geography, free limits and an experimental Pro tier. The system card adds a contemporaneous description of upload restrictions. That is enough to describe the historical release without turning later availability or a sunset notice into retroactive proof. The archive reports OpenAI’s own terms and staged status. It does not claim that Sora 2 delivered a particular commercial outcome, that every generated voice was accurate, or that app access itself settled use rights. ## Sources 1. [OpenAI: Sora 2](https://openai.com/index/sora-2/) 2. [OpenAI: Sora 2 system card](https://openai.com/index/sora-2-system-card/) ## Related reading - [OpenAI put Sora Turbo on Sora.com after its research preview](https://rangoon.ai/insights/archive-openai-sora-turbo-december-2024/) - [Runway added Gen-4 Image Turbo to its developer platform](https://rangoon.ai/insights/archive-runway-gen4-image-turbo-august-2025/) - [Midjourney opened V7 alpha testing with personalization and Draft Mode](https://rangoon.ai/insights/archive-midjourney-v7-alpha-april-2025/) ## More information - [Documentation index](https://rangoon.ai/llms.txt) - [AI access and policies](https://rangoon.ai/ai/) - [Source repository](https://github.com/hypler-dev/rangoon)