Back
Sora 2: OpenAI's Flagship Video and Audio Generation Model
webCredibility Rating
4/5
High(4)High quality. Established institution or organization with editorial oversight and accountability.
Rating inherited from publication venue: OpenAI
Relevant to AI safety discussions around increasingly capable generative models, world simulators, and deepfake/identity risks from voice and appearance cloning features; represents frontier capabilities progress rather than safety research.
Metadata
Importance: 35/100press releasenews
Summary
OpenAI releases Sora 2, a significantly improved video and audio generation model featuring enhanced physical accuracy, controllability, synchronized dialogue, and sound effects. The model represents a major step toward world simulation, better modeling physical laws including failure states, and supports injection of real-world elements like specific people into generated scenes.
Key Points
- •Sora 2 improves physical accuracy over prior models, correctly simulating physics failures (e.g., basketball rebounds) rather than 'morphing reality' to fulfill prompts.
- •Described as a potential 'GPT-3.5 moment for video,' demonstrating complex physical dynamics like gymnastics and paddleboard backflips.
- •Features synchronized audio generation including speech, background soundscapes, and sound effects with high realism.
- •Supports injection of real-world people, animals, or objects into generated environments with accurate appearance and voice replication.
- •OpenAI frames advanced video generation as critical infrastructure for training AI systems that deeply understand the physical world.
Cited by 1 page
| Page | Type | Quality |
|---|---|---|
| OpenAI | Organization | 62.0 |
Cached Content Preview
HTTP 200Fetched Feb 26, 202618 KB
Switch to
- [ChatGPT(opens in a new window)](https://chatgpt.com/?openaicom-did=6fc341e7-8e13-43c0-bda9-0246c50aff10&openaicom_referred=true)
- [Sora(opens in a new window)](https://sora.com/)
- [API Platform(opens in a new window)](https://platform.openai.com/)
Sora 2 is here \| OpenAI
September 30, 2025
[Research](https://openai.com/news/research/) [Release](https://openai.com/research/index/release/) [Product](https://openai.com/news/product-releases/)
# Sora 2 is here
Our latest video generation model is more physically accurate, realistic, and more controllable than prior systems. It also features synchronized dialogue and sound effects. Create with it in the new Sora app.
[Download the Sora app(opens in a new window)](https://apps.apple.com/app/id6744034028)
Listen to article
Share
Today we’re releasing Sora 2, our flagship video and audio generation model.
The [original Sora model](https://openai.com/index/video-generation-models-as-world-simulators/) from February 2024 was in many ways the GPT‑1 moment for video—the first time video generation started to seem like it was working, and simple behaviors like object permanence emerged from scaling up pre-training compute. Since then, the Sora team has been focused on training models with more advanced world simulation capabilities. We believe such systems will be critical for training AI models that deeply understand the physical world. A major milestone for this is mastering pre-training and post-training on large-scale video data, which are in their infancy compared to language.
Prompt: figure skater performs a triple axle with a cat on her head
With Sora 2, we are jumping straight to what we think may be the GPT‑3.5 moment for video. Sora 2 can do things that are exceptionally difficult—and in some instances outright impossible—for prior video generation models: Olympic gymnastics routines, backflips on a paddleboard that accurately model the dynamics of buoyancy and rigidity, and triple axels while a cat holds on for dear life.
Prompt: a guy does a backflip
Prior video models are overoptimistic—they will morph objects and deform reality to successfully execute upon a text prompt. For example, if a basketball player misses a shot, the ball may spontaneously teleport to the hoop. In Sora 2, if a basketball player misses a shot, it will rebound off the backboard. Interestingly, “mistakes” the model makes frequently appear to be mistakes of the internal agent that Sora 2 is implicitly modeling; though still imperfect, it is better about obeying the laws of physics compared to prior systems. This is an extremely important capability for any useful world simulator—you must be able to model failure, not just success.
The model is also a big leap forward in controllability, able to follow intricate instructions spanning multiple shots while accurately persisting world state. It excels at realistic, cinematic, and anime styles.
Prompt: Vikings Go To War — North Sea Launch (10.0s,
... (truncated, 18 KB total)Resource ID:
edc1663b7d3b8ac2 | Stable ID: ZDQ1ZTA5NT