Games & AI Video · Build & create
How AI Video Models Work
How models like Sora and Veo turn text into video: diffusion (learning to remove noise), latent compression, transformers over spacetime patches, text conditioning, why temporal consistency is hard, and costs, limits and provenance.
Step 1 of 6 · Overview30 min read ·
Key idea
Prereqs: How LLMs work; Vectors. Next: Prompting for creative tools. Sources: Ho et al., 'Denoising Diffusion Probabilistic Models' (2020); OpenAI's Sora technical report (2024); Google DeepMind's Veo documentation.Next: 1. Diffusion