OpenAI Launches Sora App to Challenge TikTok, Unveils Sora 2 Model

On Tuesday, OpenAI introduced Sora 2, a new audio and video generator that builds upon last year’s Sora model. The company also released a companion social app, also named Sora, where users can produce and share videos featuring themselves and friends through a TikTok-like algorithmic feed. Earlier reports from Wired had hinted at OpenAI's development of a new social platform.

Although we haven’t yet had access to test the invite-only app or the Sora 2 model, OpenAI has unveiled several striking examples. Notably, Sora 2 offers improved adherence to the laws of physics, enhancing the realism of generated clips. Among the public demos shared are scenes of a beach volleyball match, skateboarding moves, gymnastics sequences, and cannonball jumps from a diving board.

Sora 2 is here. pic.twitter.com/hy95wDM5nB

— OpenAI (@OpenAI) September 30, 2025

“Earlier video models tend to be overly optimistic—they might distort objects or bend reality to fulfill a text prompt,” OpenAI explained in a blog post. “In one example, if a basketball player misses a shot, the ball could instantly teleport into the hoop. But with Sora 2, that same missed shot would bounce realistically off the backboard.”

The Sora app includes an “upload yourself” option known as “cameos,” enabling users to insert themselves into any scene created by Sora. To use their own likeness in generated videos, users must submit a one-time video and audio recording that verifies their identity and captures their appearance.

This cameo feature also supports sharing, letting users authorize friends to include their likeness in generated videos—even clips with multiple people.

“We believe a social app centered on the ‘cameos’ feature offers the ideal way to experience the magic of Sora 2,” the company wrote.

The Sora iOS app is available now for download, initially launching in the U.S. and Canada, with plans to expand to more countries soon. Although the social platform currently requires an invitation, ChatGPT Pro subscribers will be able to try the Sora 2 Pro model without an invite.

Join 10k+ tech and VC leaders for growth and networking at Disrupt 2025

Netflix, Box, a16z, ElevenLabs, Wayve, Hugging Face, Elad Gil, Vinod Khosla — just a handful of the 250+ industry leaders conducting 200+ sessions designed to provide essential insights for startup growth and competitive advantage. Don’t miss the 20th anniversary of TechCrunch and this opportunity to learn from top tech experts. Secure your ticket before the event begins and save up to $444.

Join 10k+ tech and VC leaders for growth and networking at Disrupt 2025

Netflix, Box, a16z, ElevenLabs, Wayve, Hugging Face, Elad Gil, Vinod Khosla — just a few of the 250+ heavy hitters leading 200+ sessions packed with insights to drive startup growth and give you an edge. Don’t miss this chance to learn from leading voices in tech. Get your ticket before doors open and save up to $444.

Once videos are generated, they can be shared in the Sora app's feed, which appears similar to TikTok, Instagram Reels, or other short video feeds. Interestingly, Meta announced just last week a new video feed called “Vibes” in its Meta AI app (which appears to consist largely of low-effort content).

To tailor its algorithmic recommendations, OpenAI will account for a user’s Sora activity, geographic location (via IP address), past engagement with posts, and ChatGPT conversation history—though the latter can be disabled. The Sora app also includes parental controls through ChatGPT, enabling parents to disable infinite scrolling, turn off personalized algorithms, and control who can message their children. Still, the effectiveness of these tools largely depends on the parent's technical familiarity.

The Sora app will launch as a free service, which OpenAI states is intended “so people can freely explore its capabilities.” At launch, the company’s only announced monetization plan involves charging users for extra video generation during periods of high demand.

Launching a social platform demands strong user safety measures from OpenAI, which has encountered similar issues with ChatGPT. Although users can withdraw likeness access at any time, this type of feature is vulnerable to misuse. Even if a user trusts someone with their AI likeness, that person could still create misleading content that causes harm. Non-consensual videos remain a major concern with AI-generated media, creating serious harm with limited legal frameworks to regulate platform accountability.