Sora 2
OpenAI’s Sora 2 is its latest video‑and‑audio generative model, designed to produce more physically accurate, realistic, and controllable videos than earlier systems. It features synchronized speech and sound effects, more consistent world state across multiple shots, and a novel “cameos” capability that lets users insert themselves (or others) into generated scenes based on a short verification recording.
The model aims to simulate natural behaviors (for example, a missed basketball shot might bounce off the backboard rather than teleporting to the hoop) and obey physical constraints more reliably than prior video models. It’s being deployed via a social video app (“Sora”) where users can create, remix, and share short AI‑generated clips, while OpenAI also emphasizes safety, moderation, and user control over what appears in feeds or how likenesses are used.