Openai presented Sora 2, her new generative artificial intelligence model capable of creating realistic videos and audio starting from simple textual instructions. Together with this, the company has also launched an app called Sora, designed as a social platform in which users can generate personalized clips, insert themselves or their friends within scenes generated by the AI and share them in an feed that very reminds Tiktok and Instagram Reels. To access the app, you need an invitation, at least in this initial phase, but Chatgpt Pro users will have the opportunity to try the advanced Sora 2 Pro model directly without waiting for the invitation. For the moment the app is available for iPhone and can be downloaded via the App Store in the United States and Canada alone. In the future it will be gradually released all over the world.
Sora 2: more respect for the laws of physics and synchronized audio
The number one improvement of Sora 2 compared to the first version is the ability to better respect the laws of physics: the objects in videos no longer behave in an artificial or inconsistent way, but adhere more faithfully to what happens in the real world. To resume an example made by Openai in the official note with which he presented the model «If a basketball player misses a shot, the ball bounces on the scoreboard»Instead of teleporting in an unrealistic way in the basket. A fundamental step forward to obtain more credible simulations.
In the demonstration videos published by Openi, scenes such as beach volleyball matches appear, gymnastics exercises or spectacular jumps from trampolines: contexts that require rigorous respect for the dynamics of the human body and the surrounding environment.
In addition to the most accurate physics, Sora 2 is able to also generate synchronized audio: voices, environmental noises and soundtracks that make the videos more engaging. Users can choose different styles, from photojournalism to cinematographic look to anime aesthetic. The controllability has improved, with the possibility of following text instructions articulated on multiple scenes without losing visual consistency.
The social functions of the new Tiktok style Sora app
Sora’s official app then introduces a new approach to social interaction. By loading an initial registration of face and voice, each person can create their own “digital cameo” and use it within the videos generated with the AI. Others can also be authorized to include your avatar in their clips, for example by appearing in a group video. To reduce the risks, Openai has provided that anyone who maintains full control of their cameo: access can be revoked at any time, and videos that contain it are always visible and manageable by the user.
The app feed works similarly to that of other social platforms based on the sharing of short videos, albeit with some differences. The recommendation algorithm, or what decides which content to show, is enriched by a personalization system that can be “educated” with natural language. Users can therefore guide suggestions more explicitly. In addition, by default, the feed favors the videos of people followed or with which we interact, with the aim of promoting the creation of collaborative content rather than passive consumption.
Particular attention is paid to teenagers. Default limits have been set to the number of generations visible per day and introduced more stringent authorizations for the use of cameo. Through the integration with chatgpt, then, parents have parental control tools available that allow, for example, to block algorithmic customization or to regulate the settings of direct messages.
Where it will be available
On the availability front, Sora’s app was initially launched in the United States and Canada, with the intention of arriving early in other countries. It can be downloaded for free on iOS and, although access is currently available for the invitation of Openai, those who have a subscription to Chatgpt Pro can already experiment with the potential of the Sora 2 Pro model. There is also a release via API, which will allow developers to integrate Sora 2 in external applications. As for monetization, Openai said he wanted to keep the free app, limiting himself to paying extra access in moments of strong question.
