A new YouTube Short showcases a hyper-realistic rollercoaster point-of-view video that was never actually filmed—it was generated entirely by artificial intelligence. The creator shares the exact 400-word prompt used to produce the stunning first-person footage, along with a comparison of three leading AI video models: Google Veo 3.1, OpenAI's Sora 2, and Kling 2.5.
The AI-generated clip features a front-seat perspective on a massive rollercoaster, complete with realistic hand movements, trembling fingers gripping the safety bar, and intense motion blur during a vertical drop. The prompt includes detailed instructions for camera shake, lens distortion, wind noise, and even a terrified exclamation from the rider.
According to the creator, Google Veo 3.1 offers the best physical realism, Sora 2 excels at cinematic sound, and Kling 2.5 delivers faster rendering speeds. The short video has already sparked interest among AI enthusiasts and filmmakers looking to create immersive POV content without any real-world filming.
The same prompt is said to be reusable for other scenarios, with a planned underwater shark dive POV coming next.