New AI Video Generation Platform
OpenAI announced the release of Sora 2, an advanced AI model that can generate realistic video paired with synchronized audio, including dialogue in multiple languages. The system builds on the original Sora model first announced in February 2024 and released in December, offering improved fidelity to user prompts and more accurate physics simulation.
Social App Integration
The company also launched a companion social app, called Sora, that resembles TikTok with a vertical scroll and a “For You” feed. The app is currently available on iOS in the United States and Canada on an invite‑only basis, and each user receives four additional invites to share with friends. The app’s centerpiece is a feature called “Cameos,” which allows users to grant permission for the AI to generate videos using their own likeness.
Cameo Creation Process
To create a cameo, users record a short video inside the app, moving their head in specific directions and speaking a sequence of numbers. Once uploaded, the likeness can be remixed in new videos by describing the desired scene and audio in a text prompt. Users can control who may generate cameos with their likeness—options include only themselves, approved individuals, mutual connections, or everyone.
Content Limits and Moderation
OpenAI stated that the platform is designed to block the generation of X‑rated or extreme content and that public figures cannot be generated unless they have uploaded a cameo and given consent. The company also employs moderation tools to detect policy violations and copyright issues. Watermarks, moving metadata signals, and internal detection tools are embedded in every video to indicate AI generation.
Parental Controls and Safety Features
The release includes parental controls that allow a non‑personalized feed, the option to restrict direct messaging for teens, and the ability to pause an uninterrupted scrolling feed. These settings aim to give families greater oversight of the content their children encounter.
Technical Capabilities and Limits
Sora 2 currently generates 10‑second videos, with a potential upgrade to 15‑second clips for Pro users on the web, and future mobile support. Longer videos are possible but are compute‑intensive, and OpenAI is still working out how to handle them at scale.
Deepfake Risks and Safeguards
OpenAI acknowledges the risk that realistic AI‑generated videos could be used for misinformation. To mitigate this, every video includes multiple signals of AI origin, and the company has committed to not allowing free‑form text prompts for public figures in the current rollout. Nonetheless, the company noted that workarounds may emerge, echoing concerns raised by previous AI tools.
Future Outlook
OpenAI frames Sora 2 as a step toward “useful world simulators” and expects developers to gain API access soon. While the technology promises new creative possibilities, the company’s emphasis on safeguards reflects an ongoing balance between innovation and responsible use.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory SEO vous aide à automatiser le contenu d'actualités pour votre site.