## TL;DR
AI video generation exploded in 2024-2025 from research demos to commercial tools. Sora, Veo, and Runway now generate minutes of coherent video from text prompts, with professional filmmaking applications emerging.

## Core Explanation
Diffusion transformers form the architectural backbone: starting from noise, the model iteratively denoises toward a video that matches the text description. Spacetime patches treat video as 3D data (width × height × time). Training on massive video datasets teaches physics simulation: objects fall, water flows, light reflects.

## Detailed Analysis
Sora's "world model" hypothesis: training on enough video data causes the model to develop an internal understanding of physics. Veo 3's real-time mode enables interactive video generation. Runway Gen-4 offers professional video editing within existing footage. Kling (Kuaishou) and Wan (Alibaba) represent Chinese competitors.

## Further Reading
- Runway Research Blog
- Papers With Code: Video Generation
- C2PA Content Credentials Standard