Sora Update #1 (blog.samaltman.com)

Sora’s team announced two near-term changes after seeing real-world usage: they'll offer rightsholders more granular, opt-in controls over character generation (extending their existing likeness opt-in with additional constraints), and they’ll launch a revenue-sharing model to offset unexpectedly high video-generation costs. The company says many rightsholders—especially in Japan, where user engagement has been intense—are enthusiastic about “interactive fan fiction” but want explicit ability to permit, limit, or forbid uses of their characters; Sora plans to let rights holders choose and will iterate on enforcement as edge cases surface. For the AI/ML community this signals a fast-evolving production lesson: generative video platforms must pair powerful conditioning with fine-grained rights-management, real-time enforcement, and sustainable monetization. Technically this implies building consent registries, constraint layers on generation prompts or character embeddings, robust detection/filtering to catch slipped-through cases, and telemetry to understand cost per generation. The revenue-share experiment will also test incentive alignment between creators, rightsholders, and platform economics. Expect rapid iteration, policy refinement, and engineering work to scale these controls consistently across products.