Sora

Sora is OpenAI's text-to-video model, previewed in February 2024 and released to ChatGPT subscribers in December 2024. Built on a diffusion-transformer architecture, it can produce up to roughly a minute of high-resolution, largely consistent footage — a milestone that drove industry-wide discussion. The "world simulator" framing positioned it as more than a video tool, suggesting models could begin to track physics and object permanence. In the same window Google's Veo and Runway Gen-3 emerged as the other ambitious players in the space.