OpenAI Launches Sora 2 and Sora App: The Next Era of AI-Generated Video

OpenAI has doubled down on its ambitions in generative AI, specifically in video. On September 30, 2025, the company unveiled Sora 2, its next-generation audio and video model, alongside a brand new social app — Sora — built to enable users to create, share, and remix AI-generated clips in a TikTok-style feed.

Below, we break down what Sora 2 offers, explain how the Sora app works, and outline the opportunities and risks it presents.

What Is Sora 2? The Evolution of OpenAI’s Video Model

From Sora to Sora 2

OpenAI’s original Sora model, released in December 2024, allowed users to generate short video clips from text prompts with moderate realism, but with limitations in physics, consistency, and synchronized audio.

Sora 2 represents a significant leap forward, focusing on:

• Physical realism and temporal consistency: better modeling of collisions, momentum, buoyancy, gravity, and interactions so that actions appear natural.
  
• Synchronized audio and dialogue: speech, ambient sound, and effects are more closely aligned to visual motion.
  
• Steerability and stylistic control: users can specify camera moves, shot types, tempo, and visual style.
  
• Multi-shot consistency and scene coherence: characters, lighting, and dynamics remain stable across shots, reducing abrupt transitions.
  

According to OpenAI, Sora 2 “improves compliance with the laws of physics and temporal consistency.” For example, when a basketball player misses a shot, the ball bounces realistically off the rim rather than appearing inside the basket.

The Sora App: A TikTok-Style Layer over AI Video

The Sora app is the public-facing platform that makes Sora 2 accessible to creators and viewers in a familiar vertical video format.

Key Features

Vertical swipe feed
The app is designed with a TikTok-like interface: vertical videos, swipe navigation, and a personalized feed algorithm based on user activity, location, post interaction, and optionally ChatGPT history.

Cameos (self-insertion)
Users can submit a one-time verified video and audio recording of themselves. Once approved, they can authorize friends to include their likeness in generated videos. The system notifies users whenever their likeness is used, and co-ownership rules allow them to revoke access at any time.

Remix and collaboration tools
Users can remix trending videos or contribute elements such as scenes, audio, or styles to collaborative trends.

Content length and format

• Current clips are limited to 10 seconds.
  
• No uploads from the camera roll are supported; all content must be generated within the app.
  
• Audio is integrated, allowing prompts with speech or typed instructions to generate matching sound design.
  

Access and pricing

• The app launches first on iOS, invitation-only, starting in the United States and Canada. Global expansion is planned.
  
• Basic use is free, with extra charges for additional generations during high demand.
  
• ChatGPT Pro subscribers gain access to Sora 2 Pro without an invitation.
  

Why Sora 2 and the Sora App Matter

Creative opportunities

• Lower barrier to video creation: no need for cameras or editing software, just a prompt.
  
• New storytelling formats: cinematic scenes, music videos, or surreal crossovers are now within reach.
  
• Prompt engineering becomes a creative skill.
  
• Social virality: the remix and cameo features encourage collaborative trends.
  

Risks and challenges

• Misuse of likenesses: despite safeguards, a risk of deepfakes, impersonation, and non-consensual content remains.
  
• Copyright disputes: questions remain about training data and whether copyrighted works were included.
  
• Content moderation: OpenAI bans impersonations of public figures and explicit material, but enforcement may face challenges.
  
• Disinformation: realistic AI-generated clips could make it harder to distinguish between genuine and artificial media.
  

Use Cases and Early Demos

To showcase improvements, OpenAI released clips of beach volleyball, skateboarding, and diving into a pool, highlighting more natural motion and synchronized audio. Other demos include gymnastic routines, paddleboard tricks with buoyancy modeling, and figure skating with consistent character motion.

Users report fewer visual glitches, smoother frame transitions, and more natural interactions compared to the original Sora.

OpenAI’s launch of Sora 2 and the Sora app marks a significant milestone in AI video generation. By combining technical improvements in physics, sound, and consistency with a social media layer, OpenAI is moving generative video closer to mainstream use.

The challenge will be balancing creativity with ethics, ensuring that safety and privacy measures keep pace with the growing realism of AI-generated content.