Glossary Term
Generative Adversarial Network (GAN)
A neural network architecture where two models compete to produce realistic synthetic images.
A Generative Adversarial Network (GAN) consists of two neural networks — a Generator that creates images and a Discriminator that evaluates their realism. Through adversarial training, the Generator learns to produce increasingly convincing outputs. GANs powered early breakthroughs in AI photo colorization, including DeOldify and several academic research projects. While GANs can produce vibrant results, they sometimes suffer from color bleeding or instability on fine details. Newer diffusion-based models often provide more consistent quality on vintage photos with scratches, grain, and uneven exposure.