OpenAI retired the Sora 2 video-generation model on April 1, completing a shutdown sequence that began with the standalone Sora iOS app on March 24. ChatGPT no longer offers native video generation at any tier. The previously announced $1 billion Disney licensing deal was cancelled as part of the wind-down.
The full product-line shutdown is one of the clearest reversals in OpenAI’s history. Sora 2 had launched late 2025 with strong benchmarks and a TikTok-style social app. Internal economics never worked: per-generation compute ran into the millions of dollars daily, copyright complaints accumulated, and the consumer social feed cannibalised attention from ChatGPT’s core product without generating comparable subscription revenue.
The winners are the Chinese video incumbents: Kling 3.0 holds the current ELO lead, Seedance 2.0 leads on character consistency, and Vidu Q1 leads on cheap reference-to-video. Runway, Luma, and Pika continue in the professional tier. See the ai-video category for the current ranking and the Sora shutdown page for migration guidance.
Video generation in ChatGPT now requires pairing with an external tool. GPT-5.4’s native image generation via GPT Image 1.5 is unaffected.
Sources
Sources
Primary and corroborating references used for this news item.