OpenAI's next generation of AI-powered video arrives with a social twist: Sora 2 not only drives a more capable model, but also debuts with an app that makes creating clips instantaneous and, if desired, instantly shareable. The promise is simple: you record a short clip to capture your voice and face, choose the scene, and the system places you within it without losing the thread between shots.
This move provides context for a journey that began with ChatGPT and continued with image models: the first Sora already hinted at object permanence and a certain visual coherence, but now the leap focuses on physical fidelity and sound integration. The goal is for mistakes to appear human, not algorithmic glitches, and for the clip's world to retain its state when linking shots.
What's new in the Sora 2 model?
OpenAI describes Sora 2 as a significant leap forward, with more realistic behavior in everyday actions and complex scenes: if a ball doesn't go in, it bounces back where it lands and doesn't disappear; if there are multiple planes, the state of the environment is maintained. Furthermore, the system accepts richer instructions and supports short narratives with greater control over style, duration, and transitions.
Another notable improvement is the sound: the model generates voices, effects, and soundscapes synchronized with the visuals, so that dialogue blends seamlessly with the environment (from a busy street to fireworks). The result is a more complete audiovisual package that raises the bar compared to the previous version.
The social app: vertical feed, remixes, and cameos
Alongside the model, OpenAI is launching an app designed for creating and sharing content in a social context. The heart of the app is a vertical, mobile-style feed with clips up to 10 seconds long , like buttons, comments, and a remix option . The For You page is tailored using natural language and prioritizes interaction with people nearby to avoid endless passive consumption.
To appear in a clip, the app requires image and voice verification, and from there, it allows cameos in scenes generated by you or others. There's no upload from your camera roll: the capture is done within the app itself, which makes it easy to control and track permissions at all times.
Identity, well-being and security
The lookalike verification feature includes notifications whenever someone uses your image , even if the video remains in draft form. You can also revoke permissions and request the removal of clips you appear in, with human moderation to prevent harassment and impersonation.
OpenAI states that the experience aims to encourage creation over doomscrolling . For teenagers, there are viewing and generation limits , additional controls for cameos, and parental tools through ChatGPT. The company indicates that synthetic content is flagged with source signals and digital credentials to facilitate traceability.
Sora 2 availability, access, and pricing
The rollout begins on iOS by invitation only in the United States and Canada , with plans to expand access to more countries later. For now, it is not available in Spain. The service starts free with limitations based on computing power , and there will be a higher-quality Sora 2 Pro variant for ChatGPT subscribers, as well as web access and a future API.
What Sora 2 does well and what it still lacks
The examples showcase artistic gymnastics, somersaults, and action scenes with more consistent physics, and the results in animation and stop-motion look particularly solid . Even so, flaws persist: in realistic sequences, details like unnatural trajectories or bounces are noticeable (for example, in a beach volleyball match), and in linked shots, coherence can falter.
Competition and market context
Sora 2 lands in a rapidly growing competitive landscape: Runway with Gen-4 leading the creative team, Google integrating Veo 3 into YouTube, and Meta testing a feed of AI-generated short videos. TikTok, meanwhile, is tightening rules on sensitive synthetic content. OpenAI's strategy is to focus on developing its own social experience around its video engine.
Copyright and Safeguards
The company operates with filters that block potentially problematic copyright requests, which sometimes prevents the generation of certain clips. At the same time, OpenAI faces copyright lawsuits, with the New York Times case being one of the most high-profile, and defends mechanisms for content exclusion and removal for rights holders.
What the user can do
For creators, small agencies, or individual users, Sora 2 opens the door to producing clips with audio and effects quickly and at low cost. Cameo features and persistence of state between shots facilitate short, seamless pieces, while remix options and the social feed encourage collaborative , two-way collaboration .
With a combination of audiovisual model and social app, Sora 2 aims to make video creation commonplace, but with clear limits on security, identity, and rights . The platform launches with 10-second clips, mandatory image verification, and copyright filters, with currently limited availability and a Pro and API track that will shape its future development.
