openai is launching the sora app, its own tiktok competitor, alongside the sora 2 model

OpenAI Releases Sora 2 and Companion Social App

On Tuesday, OpenAI unveiled Sora 2, a new generation of its audio and video generation technology, succeeding the initial Sora model released the previous year. Accompanying this model is the launch of a dedicated social application, also named Sora. This platform enables users to create videos featuring themselves and their acquaintances, designed for sharing within an algorithmically driven feed reminiscent of TikTok.

Enhanced Realism and Physics in Sora 2

Although direct testing of the invite-only application and Sora 2 model hasn’t yet been possible, OpenAI has showcased compelling demonstrations. A key improvement in Sora 2 lies in its adherence to the principles of physics, resulting in more believable and realistic video outputs.

OpenAI’s released examples include depictions of activities such as beach volleyball matches, skateboarding maneuvers, gymnastic performances, and dives from a springboard.

Addressing Previous Limitations

“Earlier video models often exhibited unrealistic behavior, altering objects or distorting reality to fulfill a text prompt,” OpenAI explained in a blog post. “For instance, a missed basketball shot might inexplicably result in the ball teleporting into the hoop. Sora 2, however, realistically portrays a missed shot rebounding off the backboard.”

Introducing "Cameos" for Personalized Videos

The Sora app features a unique “upload yourself” function called “cameos.” This allows users to seamlessly integrate their likeness into any Sora-generated scene.

Utilizing this feature requires a one-time submission of a video and audio recording to verify identity and accurately capture the user’s appearance.

Sharing and Permissions with Friends

Users can also share their “cameos” with friends, granting them permission to incorporate the user’s likeness into videos they create, even those featuring multiple individuals.

“We believe that a social application centered around this ‘cameos’ functionality provides the optimal experience for exploring the capabilities of Sora 2,” the company stated.

Availability and Rollout

The Sora iOS application is currently available for download and will initially be accessible in the United States and Canada. OpenAI intends to rapidly expand availability to additional countries.

ChatGPT Pro subscribers should gain access to the Sora 2 Pro model without requiring an invitation, while the Sora social platform remains invite-only for now.

TikTok-Style Video Feed

Generated videos can be shared within a feed inside the Sora app, which is anticipated to function similarly to TikTok, Instagram Reels, or other platforms featuring short-form video content.

Notably, Meta recently announced the addition of a video feed, “Vibes,” to its Meta AI application.

Algorithmic Curation and Parental Controls

OpenAI will leverage a user’s Sora activity, location (determined through IP address), engagement with posts, and ChatGPT conversation history to personalize algorithmic recommendations. Users have the option to disable the use of their ChatGPT history.

The Sora app also incorporates parental controls through ChatGPT, enabling parents to set limits on infinite scrolling, disable algorithmic personalization, and manage direct messaging access for their children. The effectiveness of these controls depends on the parent’s technical proficiency.

Free Launch and Monetization Strategy

The Sora app will be launched as a free service, allowing users to freely explore its features. OpenAI plans to monetize the platform by charging users for generating additional videos during periods of high demand.

User Safety and Potential for Misuse

Launching a social platform necessitates robust user safety measures from OpenAI, which has previously faced challenges in this area with ChatGPT.

While users can revoke access to their likeness at any time, the potential for misuse remains. Even trusted individuals with access to a user’s AI likeness could generate deceptive content with harmful intent.

The creation of non-consensual videos is a recurring issue with AI-generated video, causing significant damage with limited legal frameworks governing platform accountability.