.png){kind=logo}
OpenAI has announced the launch of Sora 2, a next-generation text-to-video AI model. The company claims that it is capable of performing ‘exceptionally challenging, or even impossible,’ feats that earlier video generation models could not achieve.
Sora 2 introduces previously unseen realism to AI-generated video, which accurately replicates physical laws, according to the company. For instance, whereas older models would teleport a basketball into the hoop after a shot had missed, Sora 2 causes the ball to bounce naturally off the backboard.
In addition to physical realism, the model is superior in controllability. OpenAI states it can execute complicated multi-shot instructions with consistency in the world it creates. The AI is capable of producing videos in realistic and anime appearances. It can also generate background soundscapes, dialogues, and sound effects synchronized with action. Sora 2 can uniquely put humans or other real objects into scenes it generates, enabling possibilities for highly personalized content.
Sora 2 enters a crowded field, competing with Google Veo-3, which was widely publicized earlier this year. While both create videos with audio synchronization, OpenAI highlights the model’s more advanced physics simulation and subtler control.
The firm frames it as a step toward completely simulating reality in AI-generated video, recognizing that, although far from ideal, the model demonstrates a significant advance in neural network scaling on video data.
In addition to the AI, OpenAI is releasing a stand-alone iOS social media app named Sora. The app enables users to create, remix, and discover AI-created videos in a feed related to their interests.
A highlight feature is Cameos, where users can record a short snippet to become part of AI-created scenes made by themselves or others. Developed for creative collaboration, the app provides tools that allow users to bring themselves and their friends into an interactive virtual world.
OpenAI is working towards minimizing addictive scrolling by offering users control over the algorithm of content. The feed can be directed via natural language, and default content will be videos from the accounts people are following or having contact with. Sora’s algorithm also gives priority to clips that have the potential to inspire new creation, not mere passive consumption.
With software like Sora 2 and its companion application, OpenAI combines state-of-the-art video creation with an element of social interaction. The tech giant believes that the release of this product marks a time when AI-generated content goes mainstream.
Cinematic experiences can be created by users and even allow them to immerse themselves in virtual worlds. Although the technology is still in development, it represents a significant leap forward for AI-generated digital media.
