Glossary Term
AI Diffusion Models
Generative AI models that gradually add detail to noise until a full color image emerges.
AI diffusion models are a class of generative neural networks that learn to reverse a noise-addition process. Starting from random noise, the model iteratively denoises the input until a coherent image appears. In photo colorization, diffusion models analyze luminance patterns, edges, and semantic regions in a black-and-white source image, then predict plausible chrominance (color) values for each pixel region. Compared to older GAN-based colorizers, modern diffusion models produce finer texture detail and more stable skin tones — especially on vintage portraits with film grain and fading.