On September 30th, OpenAI unveiled its next-generation video generation model, Sora 2, alongside a new iOS social app also named “Sora.”
Compared to Sora 1 released in February this year, Sora 2 demonstrates significant improvements in physical accuracy, realism, and controllability. The new model can generate complex scenes including Olympic gymnastics routines and water sports while accurately simulating physical laws. For instance, when a basketball shot misses, the ball now realistically bounces off the backboard instead of “teleporting” to the hoop as in previous models.
Sora 2 also supports synchronized dialogue and sound effects generation, capable of creating complex background soundscapes, voice, and audio effects. The model can follow complex instructions across multiple shots while accurately maintaining scene consistency.
The newly launched Sora app adopts a TikTok-style short video format, allowing users to create and remix videos while discovering new content through a personalized feed. The app’s core feature is…