Introduction
Diffusion models power most modern AI image and video generators.
What is a Diffusion Model?
Diffusion models work by gradually adding noise to data, then reversing the process.
Training Process
Billions of image-text pairs train the model to understand visual concepts.
Inference (Generation)
The model starts with pure noise and iteratively denoises it guided by the prompt.
Latent Diffusion
Most models operate in latent space for faster generation.
Why Video is Harder
Video requires maintaining consistency across frames.
Conclusion
Diffusion models represent a breakthrough in AI creativity.