Diffusion models and, more generally, Flow Matching are the state-of-the-art in visual generative AI. Their derivation is not straightforward. Here's what they are and how to use them - abridged.