OpenAI has officially launched Sora 2, its most advanced video and audio generation model to date, alongside a new social app that could rival TikTok — a move that puts the company directly into the global short-form video market.

The company describes Sora 2 as the “GPT-3.5 moment for video,” a leap forward from the original February 2024 release, which it had framed as the “GPT-1 moment” when video generation first began to show simple behaviors like object permanence. With Sora 2, OpenAI says the model can now handle tasks previously considered impossible, combining world simulation, controllability, and synchronized audio in ways earlier systems could not.

A key advancement is faithful physics modeling. Earlier video models often distorted reality to fulfill a prompt, producing results that looked unnatural — like a golf club vanishing mid-swing or a basketball teleporting into the hoop. By contrast, Sora 2 captures the messier but more realistic outcomes of the physical world, such as a basketball bouncing off the backboard instead of always scoring.

“This is an extremely important capability for any useful world simulator—you must be able to model failure, not just success,” OpenAI wrote in its release.

Building on this foundation, the model can now generate Olympic-level gymnastics routines, paddleboard tricks that reflect buoyancy and rigidity, and complex multi-shot sequences while maintaining consistency in the surrounding environment.

Sora 2 also delivers a major improvement in creative control. It can follow intricate multi-shot instructions, preserve world state across edits, and produce outputs in multiple styles — from realistic and cinematic to anime-inspired visuals. Beyond video, the model integrates synchronized dialogue, sound effects, and background audio, creating more immersive and lifelike results.

One of the most distinctive features of the new Sora app is cameos, which allow users to insert themselves directly into AI-generated videos. To activate the feature, users provide a one-time video and audio recording that verifies their identity and captures both appearance and voice.

Once enabled, cameos can be shared with friends, allowing others to include a user’s likeness in collaborative AI-generated clips. The app supports multi-person scenarios, enabling groups of friends to appear together in shared videos. It also allows insertion of any human, animal, or object into a scene. In one example, OpenAI demonstrated an Olympic figure skater performing a routine while a cat sat balanced on her head throughout the performance.

OpenAI framed cameos as a “new and unique way to communicate with people,” likening the feature to the progression from text messages to emojis to voice notes. The company emphasized that the app is “made to be used with your friends,” framing cameos as a social rather than solitary experience.

However, TechCrunch highlighted potential risks, noting that even if a user trusts someone with access to their cameo, that person could still generate deceptive or harmful content. To mitigate this, OpenAI said cameo permissions are fully revocable at any time, giving users control over where and how their likeness appears.

The Sora iOS app is a social platform with a customizable short-form video feed where users can create, remix, and share AI-generated videos. Designed in a format reminiscent of TikTok and Instagram Reels, it gives OpenAI a direct entry into the competitive short-form social video market.

At its core, the app is built to be creativity-first rather than engagement-maximized. By default, the feed is heavily biased toward content from people a user follows or interacts with, and toward videos that the model predicts will inspire further creation. OpenAI said it is “not optimizing for time spent,” a deliberate choice to avoid addictive scrolling patterns.

To support this, the company has developed a new class of recommender algorithms that can be instructed through natural language, allowing users to fine-tune what they see. The app also includes periodic wellbeing polls, prompting users to adjust their feed proactively if needed. For example, users could ask to see more of their friends’ videos, fewer remixes, or more cinematic-style clips.

At a time when major platforms are moving away from the social graph, OpenAI said cameos are designed to reinforce community. In internal testing, the feature helped colleagues form new friendships, pointing to how generative AI could reshape the way people connect socially.

OpenAI has emphasized that safety and wellbeing are central to the Sora app’s design, acknowledging long-standing concerns around social media and generative AI. Platforms like TikTok and Instagram have faced criticism for encouraging doomscrolling, addictive engagement loops, and algorithmic optimization for screen time. Generative video adds new risks, including non-consensual content generation, impersonation, and bullying. OpenAI said it aimed to address these issues at launch rather than as an afterthought. You can find their feed philosophy here.

To mitigate risks, the company is introducing a range of protective measures:

In addition to moderation, OpenAI is giving users end-to-end control over their likeness. Cameo permissions can be granted or revoked at any time, and users can view and remove any video containing their likeness — including drafts created by others. The company said these protections are intended to set a higher standard for consent and provenance in AI video. You can view their Sora 2 Safety Doc here.

On the business side, OpenAI is rejecting ad-driven monetization and engagement-maximized incentives that dominate other social platforms. Instead, Sora will be free with generous limits, charging only for additional video generations when demand is especially high. OpenAI said this approach is designed to encourage exploration without creating pressure for longer screen time.

The Sora iOS app is available to download now, with the initial rollout beginning in the U.S. and Canada. Users can sign up in-app for a notification when access opens for their account, and OpenAI says it plans to expand availability to additional countries soon.

After receiving an invite, users will be able to access Sora both through the iOS app and via sora.com, where the model is integrated into the platform’s video library.

In addition, ChatGPT Pro subscribers can experiment with the higher-quality Sora 2 Pro model on sora.com — and soon within the Sora app itself. OpenAI also plans to make Sora 2 available through its API, while continuing to support the earlier Sora 1 Turbo model.

Q: What is Sora 2?
A: Sora 2 is OpenAI’s new video and audio generation model, capable of realistic physics, synchronized sound, and multi-shot control.

Q: What is the Sora app?
A: The Sora app is a new invite-only iOS social platform where users can create, remix, and share AI videos in a TikTok-style feed.

Q: How do cameos work?
A: Cameos let users upload a short verification video to insert themselves into AI-generated content and optionally share their likeness with friends.

Q: Is Sora free to use?
A: At launch, Sora is free with usage limits. OpenAI may charge for extra video generations during high demand.

Q: Where is Sora available?
A: Sora is launching first in the U.S. and Canada, with plans to expand. Users can also access it at sora.com.

The release of Sora 2 is more than just another AI upgrade — it represents a step toward AI systems that can simulate the physical world. That matters because realism isn’t just cosmetic: it’s what could eventually make AI capable of powering robotics, training agents in virtual environments, or building next-generation entertainment. By showing it can handle the “messiness” of reality, Sora 2 pushes AI closer to applications far beyond social media.

At the same time, the launch of the Sora app places OpenAI in a new role — not just as a research lab, but as a consumer platform operator. That shift carries real implications. If Sora takes off, OpenAI will be competing directly with companies like TikTok, Instagram, and YouTube for people’s time and creative output. Success would mean that generative AI stops being a niche tool and starts becoming part of everyday digital culture, woven into how people create and communicate.

But the stakes are high. Generative video brings risks of misuse, manipulation, and non-consensual content, and OpenAI will be tested on whether its safety-first approach can actually hold up at scale. If it works, Sora could become a model for how AI platforms can be designed to prioritize wellbeing rather than maximize engagement. If it doesn’t, it could face the same backlash that other social platforms have endured.

For users, this moment signals a shift: AI video isn’t a futuristic demo anymore — it’s becoming a social experience you can download to your phone today. That immediacy is why Sora matters. It’s not just about better AI models; it’s about whether AI will reshape how people connect, create, and spend their time online.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.