- Sora 2 delivers synchronized video and audio with greater physical realism and multi-shot continuity.
- The iOS social app powers verified cameos, remixes, and a customizable feed.
- Available by invitation in the US/Canada; in Spain, there are methods for installing and using it with a VPN.
The tech conversation of the last few weeks has revolved around Sora 2, OpenAI's new video and audio model and a social application that aims to make audiovisual creation as seamless as writing text. Beyond the hype, we're talking about a proposal that integrates image, movement, and sound generation from a single command, with a social layer designed to create and remix clips in the style of short-form video platforms.
In the following lines, you'll find out what Sora 2 is, how it works and how it improves upon Sora 1 , its rollout status (including its availability in Spain), what its new iPhone app offers, how to configure cameos with your face and voice, and what limitations and security measures have been announced. We'll also review access methods, from the US/Canadian App Store to using a VPN on the web and mentions of third-party integrations.
What is Sora 2 and how does it work?
Sora 2 is both a machine learning model and a product layer . The model converts text descriptions—and optionally source images—into short videos with synchronized sound; the product portion includes a website and, most importantly, a social iOS app where you create, share, and remix content.
The idea is familiar if you're used to ChatGPT or DALL·E: you type what you want and the AI interprets it in natural language . Where Sora 2 stands out is that it produces the visual clip and audio in the same process, so lip-sync and sound effects associated with what you see (footsteps, slamming doors, ambient sounds) are already perfectly aligned.
The technical foundation boasts a more realistic world simulation : greater object permanence, coherent relationships between elements within the scene, and adherence to plausible physical dynamics. This translates into more natural movements and fewer strange artifacts compared to the first generation.
In addition to starting from scratch with text, the system allows you to begin with an image and generate an animated scene from it. And, with the cameo feature, you can insert your own face and verified voice to star in the video, opening the door to personalized clips without the need for traditional filming.
New features and improvements compared to Sora 1
The previous version was already impressive in terms of the quality of individual frames , but it failed to maintain consistency and realistic physics in motion. Sora 2 takes a leap forward on three key fronts: motion realism, seamless transitions between shots, and synchronized native audio.
In terms of realism and physics, the model better captures momentum, collisions, buoyancy, and friction , so actions like jumping, a backflip on a paddleboard, or a basketball bounce feel believable. Where Sora 1 could "teleport" objects or break trajectories, the new version more closely approximates real-world behavior.
In terms of multi-shot consistency, Sora 2 maintains objects, characters, and styles across multiple shots , supports complex commands, and executes camera movements and framing with greater fidelity. This facilitates sequences with small plots and transitions without losing key elements along the way.
Audio is now natively integrated: dialogue, Foley effects, and ambient sounds are generated alongside the video. You can include a script in your prompt so a character "speaks" and the clip plays with lip-syncing, along with soundscapes and effects positioned in time as specified.
Finally, the stylistic range is broader. You can request photorealism, cinematic style, anime aesthetics , or other variations, and the system better understands these nuances to match the aesthetic to your description.
Cameos: Your face and voice in videos
One of the standout features is Cameos. In the app, you set up your cameo with a quick verification process: reading numbers and turning your head to capture facial features and voice. With this material, Sora 2 can insert you as the protagonist of generated scenes, without training complex models or using LoRAs.
The tool includes consent controls : you choose whether it's for you alone, approved friends, your entire friends list, or the whole community. You can also revoke access later or remove videos that include you, which is vital to prevent unwanted use.
The social phenomenon quickly emerged. Sam Altman's public cameo sparked a wave of memes with humorous remixes; the platform allows a single generation to contain multiple cameos, so you can mix with other people in the same scene for collaborative sketches.
On a practical level, generating a clip takes a few minutes , you can launch several at once, and there are daily limits (up to 100 per day are cited in user experiences). For now, typical clips are around nine seconds long, and the format can be vertical or panoramic depending on what you specify in the prompt.
OpenAI complements the model with a social app for iPhone where the feed is reminiscent of TikTok: a carousel of videos, comments, likes, and native remixes. Unlike other networks, the default description displays the original prompt, making it easier to learn and replicate ideas.
The app includes a "mood" selector to adjust what you feel like watching , and OpenAI claims its recommendation algorithms can be configured using natural language. To prevent doomscrolling, the company says it prioritizes content creation over consumption, with regular well-being surveys and options to customize the feed.
In tone and spirit, many users describe the experience as a return to the agility of Vine : quick clips, straightforward humor, and less pressure for perfection. There are also mechanisms for remixing other people's videos and expanding memes with variations, which speeds up the creative cycle.
However, frictions are emerging regarding intellectual property. The app directly blocks the creation of copyrighted characters (The Simpsons, Dragon Ball, etc.), although some users attempt to circumvent filters with alternative words. The line between inspired styles and the use of copyrighted characters is under scrutiny, and the company is reinforcing moderation and limits.
Availability, countries and what's happening in Spain
At launch, Sora 2 is in an invitation-only beta phase and initially available in the United States and Canada. There is no official release date for Spain, nor for the final version that will be available without an invitation.
The Sora app for iOS is currently only available in the US and Canadian App Stores. A web version exists, but certain popular features—like setting up cameos—require the iPhone app, at least initially.
The service launches for free but with varying limits depending on demand (available computing capacity). If you have a ChatGPT paid plan, you'll have priority, higher image quality, and fewer restrictions, although these details may change as the rollout expands.
How to download the app on iPhone if you're in Spain
There's a way to install the app without a VPN if you're in Spain: use a US or Canadian Apple account . You won't lose your data and you can switch back to your regular account after downloading; however, you'll have to do updates with the US account.
- Go to the Apple ID website (appleid.apple.com/account) and Create an account with your real data.
- In the Country or Region field, choose United States or Canada.
- Complete the captcha and creation ends.
- On your iPhone, open the App Store, sign out, and log in with the new account. Search for “Sora by OpenAI” and download.
After installing, you can switch back to your Spanish account. The app will continue to work ; when it's time to update, you'll need to switch accounts again. Keep in mind that you'll need an active invitation to play Sora 2 within the app.
If you prefer a browser, the web version of Sora also requires an invitation, and users outside the launch countries must use a VPN with a server in the US or Canada. Without that VPN, access to the first version of the model is reportedly limited.
Log in with your OpenAI/ChatGPT account and check which features are available. For now, cameos are configured in the iOS app , so key pieces of the social experience are missing from the web version—something that could change as the rollout is completed.
Prices, limits and the mention of “Sora 2 Pro”
OpenAI positions Sora 2 as a service with limited free use and premium options tied to its subscription plans. A Sora 2 Pro variant has been mentioned for creators seeking higher resolution and longer video durations, integrated with ChatGPT Pro, and with a paid API in development.
The sources analyzed mention priority access for subscribers and a future API with billing per clip or second, though no official rates were detailed at the time of the announcement. They also cite references to ChatGPT Pro as an experimental access point, as the web app will be integrated with this ecosystem.
Aside from the official channel, some third parties—like CometAPI—claim Sora 2 compatibility via API under a pay-as-you-go model (mentioning $0,16 per execution with streaming) and show examples of chat completion requests against a unified endpoint. If you're a developer, always review terms and limitations before integrating third-party services.
curl --location --request POST 'https://api.cometapi.com/v1/chat/completions' \
--header 'Authorization: sk-XXXX' \
--header 'Content-Type: application/json' \
--data-raw '{
"model": "sora-2",
"stream": true,
"messages": [{"role": "user", "content": "Generate a cute kitten sitting on a cloud, cartoon style"}]
}'
In terms of performance, there are testimonials that cite clips of up to 10 seconds with perfect audio-video synchronization and improvements over alternatives like Veo 3 in certain scenes. These figures may vary depending on availability, cost control, and model evolution.
Limitations, security and intellectual property
OpenAI has incorporated security and moderation controls : default limits for teenagers, tools to adjust personalization, harassment detection, human moderation, and parental control options from the ChatGPT settings (usage times, disable personalization, moderate direct messages).
The company claims to have mechanisms in place to restrict public figures and block explicit content . They also mention liveness checks for cameos, metadata and watermarks, and options to revoke the use of your image, all aimed at preventing impersonation and misuse.
In terms of copyright, the app attempts to filter out characters with protected rights ; even so, some users find loopholes using alternative terms. The issue is being monitored by rights holders, and OpenAI is already facing litigation for training with copyrighted content, so constant adjustments are to be expected.
Sora 2 should not be used where unequivocal factual accuracy is required . It is a generative system: it can invent plausible details, so in sensitive information contexts or with legal requirements, it should be treated as a creative support tool and not as a substitute.
User experience and remix culture
The initial feed conveys a sense of lightness and humor : carefully crafted cuts designed for comedic effect, intentionally awkward silences, and glances at the camera that the model adds even without being explicitly asked, if they contribute to the humor. This fine-tuning enhances the viral potential of the short sketches.
The app facilitates creative imitation with remix features and prompts visible by default. This accelerates collective learning and encourages ideas to spread rapidly with minimal variations, something we've already seen succeed in other short-form video formats.
Some describe Sora as a jab at Meta , which launched Vibes, its network of self-generated clips with a creative/artistic focus. OpenAI's approach is more irreverent and accessible, aiming to prioritize creation over passive consumption to combat endless scrolling fatigue.
Competition and panorama
The industry is moving fast. Google is pushing forward with Veo 3 and creator tools on YouTube; Meta is advancing with Vibes; TikTok maintains more restrictive policies on AI-generated videos dealing with sensitive topics. The battle may also extend to monetization, as companies strive to best integrate these features into real-world workflows.
In parallel, there is an ongoing debate about training with existing content (TV programs, videos from streaming platforms, etc.). Some publications suggest that even articles like this one could be used to train future models, emphasizing agreements, licensing, and transparency.
Within the Apple ecosystem, related resources such as Image Playground for generating images have been cited , along with more general content about apps and games that help to define the creative context on iOS. The growing role of YouTube within Google's search engine itself is also discussed, showing how video is becoming a central element of discovery.
Ideal use cases… and when not to use them
- Short-form social video: rapid iteration, remix culture, and cameos for viral clips.
- Prototyping for film, advertising or games: Visual and animatic mockups with coherent audio.
- Educational and marketing animations: narration aligned with visual elements and sound.
- Small studios and creators: polished results without big budgets.
- Large format and precision VFX: frame-by-frame control even superior to traditional pipelines.
- Contexts that require factual accuracy: The model may invent plausible but erroneous details.
Looking at the whole picture, Sora 2 brings the idea of a finished video clip closer than ever before : describe the scene, add style, insert your cameo, and get a video with synchronized sound ready to share or remix. There's still a lot to refine—from global availability to licensing and costs—but the creative and social potential already evident in this early stage explains why so many people can't stop trying it.