OpenAI has recently unveiled its latest video generation model, Sora 2.0, during a media tour of its data center in Abilene, Texas. This new iteration builds upon an earlier version released as a research preview earlier this year. OpenAI claims that Sora 2.0 significantly enhances physical realism, audio-video synchronization, and multi-shot storytelling capabilities.
Accompanying the model is a new iOS application, also named Sora, which is currently in an invite-only phase. This app allows users to create, remix, and appear in AI-generated videos using text or images. It features built-in controls aimed at ensuring safety, likeness, and provenance in content creation.
The launch has stirred discussions regarding intellectual property rights, as reported by The Wall Street Journal. OpenAI is informing studios and talent agencies that unless they opt out, their copyrighted materials may be incorporated into Sora-generated content. The company asserts that its approach aligns with existing content policies utilized in its image generation tools and encourages fan expression through the remixing of familiar fictional worlds.
Executives at OpenAI have noted high user interest in leveraging AI for storytelling and character interaction, suggesting that Sora offers new avenues for engagement with popular fictional universes. Furthermore, OpenAI has established a distinction between copyrighted intellectual property and individual likenesses. Individuals cannot be generated without prior verified consent, allowing them to maintain control over their appearances in AI-generated content.
The Sora app is set to roll out in the U.S. and Canada, with plans for a Pro version, web access, and a developer API in the near future.