OpenAI to shut down Sora app months after launch as focus shifts to agents

OpenAI plans to shut down its Sora AI video generator just months after launching a standalone app, as the company moves to simplify its product lineup and redirect resources toward core AI systems.

The decision also ends OpenAI’s partnership with The Walt Disney Company, which had licensed characters such as Mickey Mouse and Cinderella for use in Sora and included a $1 billion equity-based investment structure tied to the project.

Sora launched in late September as a consumer-facing app that allowed users to generate short AI videos from text prompts, remix content, and share clips in a social feed. The app quickly climbed the Apple App Store rankings before losing traction in the following months.

OpenAI will also discontinue Sora’s API, cutting off developer access to the video generation system. The company said it will provide timelines for the shutdown and guidance on preserving user-created content.

The move comes as OpenAI shifts focus toward AI agents and a new model internally referred to as Spud, which CEO Sam Altman expects to release in the coming weeks. The company is also building a unified desktop platform that combines ChatGPT, coding tools, and browsing capabilities.

Compute constraints appear to be a key factor. AI video generation requires significantly more processing power than text or image models, and OpenAI is prioritizing infrastructure and systems that can scale across broader use cases.

Sora’s underlying technology will continue as part of OpenAI’s research into world simulation, which the company sees as foundational for robotics and real-world task automation. An OpenAI spokesperson said the research team will focus on advancing systems that interact with physical environments.

The shutdown follows growing concerns around AI-generated video, including risks tied to deepfakes and misinformation. While Sora attracted early interest, it faced challenges in sustaining engagement and balancing safety considerations at scale.