OpenAI Launches Sora 2 and a Social App for Physically Realistic AI Video
OpenAI launched Sora 2, a video-and-audio generation model that is more physically accurate, realistic, and controllable, with synchronized dialogue and sound effects. A new invite-based Sora iOS app emphasizes creation, remixing, and "cameos" that let users insert their verified likeness into scenes, with strong controls over consent and wellbeing. The rollout begins in the U.S. and Canada with generous free limits, Sora 2 Pro for ChatGPT Pro users on sora.com, and an API planned.
Key Points
- Sora 2 substantially advances physical realism, world-state persistence, and controllability in video-and-audio generation, including synchronized speech and sound effects.
- The new Sora iOS social app focuses on creation and remixing, with a customizable feed and "cameos" that let users inject their verified likeness into generated scenes.
- Safety and wellbeing are central: instructable recommender controls, bias toward your social graph, limits for teens, parental controls, human moderation, and robust consent over likeness.
- Availability starts in the U.S. and Canada with generous free limits; Sora 2 Pro is accessible to ChatGPT Pro users on sora.com, Sora 1 Turbo remains, and an API is planned.
- OpenAI positions Sora 2 as a GPT‑3.5‑level milestone for video, moving toward general-purpose world simulators while acknowledging current imperfections.
Sentiment
The overall sentiment is skeptical and concerned. While commenters acknowledged the technical achievements, the prevailing reaction was negative—questioning OpenAI's strategic motives, expressing existential concern about the societal implications of realistic AI-generated video, and noting that the outputs still fall short of convincing realism.
In Agreement
- The data race between OpenAI and Google is significant, implicitly acknowledging Sora 2 as a meaningful competitive product
- Multiple commenters acknowledged the technical impressiveness of the video generation capabilities
Opposed
- OpenAI is turning Sora into a consumption-focused social network ("AITok") rather than a genuine creative tool, with minimal editing controls and no long-form video support
- AI video generation represents a dangerous moment for society, enabling deepfakes and eroding trust in all visual media
- The generated videos remain deeply uncanny, with faces and mouth movements looking noticeably artificial and even nauseating to watch