OpenAI has released Sora 2, its next-generation video model, alongside a new TikTok-style social app that allows users to create and star in AI-generated videos, confirming an earlier scoop from Wired. The invite-only iOS app, also named Sora, is built around a feature that lets users insert their own likeness into shareable video clips.
- Ready for your close-up: The app’s centerpiece is ‘cameos,’ where users upload a short video to capture their likeness for use in generated scenes. After verifying their identity, users can grant friends permission to include them in other video creations. According to The Verge, the person whose likeness is used is considered a "co-owner" of the video and can delete it at any time.
- A leap in realism: The company is framing the release as a potential “ChatGPT moment for video,” claiming Sora 2 is better at obeying the laws of physics than prior models. In a blog post, OpenAI stated, "if a basketball player misses a shot, it will rebound off the backboard." The model also marks a first for OpenAI, as it can generate synchronized audio and speech directly from a text prompt.
With this launch, OpenAI isn't just releasing a new tool; it's stepping into a crowded arena, setting up a direct clash with Meta's "Vibes" feed and Google's YouTube-integrated video models for the future of user-generated content.
The move immediately raises questions in Hollywood about the future of professional content creation. The launch also comes as OpenAI is fighting a series of high-profile copyright lawsuits, including one from The New York Times, over allegations it trained its models on copyrighted material. Meanwhile, the U.S. tech giants face intense international competition from Chinese alternatives like Kling and Hailuo that are also pushing the boundaries of AI video.
