Sora 2: OpenAI’s New Video-Creating Model and App

OpenAI has introduced Sora 2—a new AI model for generating videos and audio, along with a social app of the same name. Here’s what we know about the release date, how to get an invite code, and how to download the app.

What is Sora 2?

In late September 2025, OpenAI finally unveiled Sora 2—their latest model for creating videos from text prompts.Sora 2 understands how the real world works. The AI handles complex movements with greater confidence—for example, gymnastics routines or water sports, where every detail matters.Another impressive feature is synchronous audio generation. The AI automatically selects music, adds sound effects, and even voices dialogues that perfectly match the characters’ lip movements.It’s also worth mentioning the new Sora app for iPhone. It operates like a familiar feed, similar to TikTok, but is fully dedicated to AI-generated videos.There, you can not only scroll and watch, but also easily create your own clips or remixes of others’. By the way, AI-generated videos are already going viral on mainstream social media platforms.But the most exciting feature is “Cameo.” It lets you “insert” yourself or friends into any generated video. Just record a short video and voice sample once for verification—and then you can appear in any scenes, from fantasy worlds to space operas.That said, access is currently limited. The app works via invites and is only available in select countries, plus servers sometimes struggle with high demand. But those who’ve tried it say it looks incredibly convincing. For more details, check out our Sora review.

Key Innovations in OpenAI Sora 2

• Video and audio in one generation: This is no longer just moving images. The model creates a complete clip where the audio track—dialogue, sounds, music—is generated simultaneously with the visuals and perfectly synchronized. For example, you can get a scene with dialogue where lip movements match the spoken words, all in one click, without editing.
• Physical consistency and realism: Significant work has been done here on understanding physics. If in older models a ball could magically teleport into the basket, in Sora 2 it will realistically bounce off the rim on a miss. Movements are now smoother and more natural, especially in complex scenes like gymnastics or water sports.
• Improved prompt interpretation: The model better understands complex, multi-part requests. You can describe not just a static scene, but a full sequence of actions, shot changes, or angles, and Sora 2 will faithfully execute it while maintaining consistency throughout. You can even use cinematic language, specifying shot types (e.g., “close-up,” “side view”) and camera movements (“zoom in,” “pan”).
• Handling diverse styles: Sora 2 confidently works with a wide range of aesthetics—from hyperrealism to anime and various artistic styles. The model maintains quality and “sticks to” the chosen visual language throughout the clip. Built-in presets—ready-made setting templates—help establish the desired atmosphere right away.

What to Know Before Using It

• The “Cameo” feature—it’s available, but optional. Users must complete verification by recording a short video and voice sample so the system can “remember” them. Afterward, you can realistically insert your face into any AI generations.
• Strict filters—yes, OpenAI has implemented tight content restrictions, especially regarding well-known brands and intellectual property. Many users encounter generation refusals, and policies may evolve over time.
• Access—the model is currently distributed via invitations, with priority given to ChatGPT Pro users and some active participants from the first Sora version. For others, including Plus subscribers, there’s a waitlist in place.
• Generation time—yes, during peak loads, creating a clip might take a few minutes.

What We Know About Availability and How to Download Sora 2

At the moment, you can’t fully use the Sora 2 app yet, as it’s available to users in the US and Canada. The release date for Sora 2 was September 30, 2025.You can gain access in the following ways:

• App and web access: The main method is through the iOS app “Sora” or the website sora.com.
• Sora 2 invite code: Access is handled through an invite system. Each new user can invite a few friends.
• Subscription plans: Free usage has limits. For serious work, the experimental Sora 2 Pro model is available via a ChatGPT Pro subscription.

Conclusion

What do we know about Sora 2 so far? The tool excels at dialog scenes, simple everyday clips, and basic ads—its capabilities suffice for such content. Free access will likely lead to a flood of entertaining and promotional videos created with this AI across the web.However, there are limitations. For dynamic action sequences, the quality is average—sometimes it feels like the “low-budget” staging reminiscent of 1990s B-movies. With stylization, fantasy, and complex artistic tasks, Sora 2 still falls short. Outside short-video social media formats, results often look awkward.At present, the tool isn’t suited for professional production—it’s more for mass entertainment. You can make short funny sketches, but for more complex projects, it might lack the depth. For basic ads, Sora 2 works well in principle, but there’s a key restriction: the system doesn’t generate animations of realistic people or maintain character consistency from frame to frame.As a free alternative to Veo, it’s a solid choice. But it can’t be called a “Veo killer,” let alone a replacement for models like Kling, Hailuo, or Seedance. It’s a unique tool with its own strengths and weaknesses.