Sora 2

The webpage you linked announces OpenAI’s latest advancement in AI-generated media: Sora 2, a text-to-video and audio model that’s a big step up from its predecessor.

What is Sora 2?

Sora 2 is described as OpenAI’s flagship system for generating videos and audio, emphasizing realism and control. It’s compared to the “GPT-3.5 moment for video,” indicating it’s at a mature development stage but still evolving.

Key Features and Capabilities

• Physical Accuracy and Realism: It handles complex dynamics like buoyancy, rigidity, and failures (e.g., a basketball missing the rim and rebounding properly, without unnatural morphing). Examples include backflips on paddleboards, Olympic gymnastics, or a figure skater landing a triple axel with a cat balanced on her head.
• Audio Integration: Generates synchronized dialogue, sound effects, and immersive background audio with high fidelity.
• Styles and Persistence: Creates cinematic, realistic, or anime-style videos that follow multi-shot instructions while maintaining consistent world states.
• Cameos Feature: Inserts real-world elements like a person’s face and voice (or even animals/objects) into generated scenes.

Improvements Over Sora 1

The original Sora (launched in February 2024 and called the “GPT-1 moment for video”) had issues with physics-defying morphs and less accurate simulations. Sora 2 improves through better pre- and post-training on vast video datasets, resulting in more physically grounded outputs that obey real-world laws.

---

Availability and Access

• Launch Details: Rolled out via a new iOS app called “Sora” on September 30, 2025, starting in the U.S. and Canada, with expansion planned soon. Web access is at https://sora.com (invite-required).
• Free Tier: Available for free with generous usage limits, though compute constraints apply. Potential future pay-for-extra option if demand spikes.
• Pro Version: ChatGPT Pro subscribers get an experimental higher-quality “Sora 2 Pro” on the web (app coming soon).
• Legacy Support: Sora 1 Turbo remains usable, and old creations persist in user libraries.
• Future Plans: API access is in the works.

Demos and Examples

The page showcases prompts and outputs like:

• Vikings battling in the North Sea.
• Mountain explorers shouting in snowy conditions.
• A detailed dragon scene: “A massive dragon with iridescent scales flies low over jagged ice spires in a frozen tundra at dawn, its wings casting thunderous shadows as glaciers calve with booming cracks in the distance—4K, 50mm lens, epic scale with tactile realism.”

Integrations and Related Info

• Integrates with OpenAI’s language models for app features like personalized feeds via natural language.
• Parental controls powered by ChatGPT for managing app usage.
• Additional docs on feed philosophy and safety: https://openai.com/index/sora-feed-philosophy/ and https://openai.com/index/launching-sora-responsibly/.

If you have a specific question about Sora 2 or want me to dive deeper into any aspect, let me know!