What Is a GAN (Generative Adversarial Network)?
An earlier approach to AI image generation: two networks locked in competition, one creating fakes and one trying to catch them, each getting better because of the other.
Two networks, competing
A generative adversarial network (GAN) pairs two networks trained against each other: a generator that creates fake content, and a discriminator that tries to tell real examples from the generator's fakes. Each one improves by trying to beat the other — the generator gets better at fooling the discriminator, and the discriminator gets better at catching increasingly convincing fakes.
Why the competition works
This adversarial setup gives the generator a moving, ever-improving target to satisfy, instead of a fixed rule for what "realistic" means. As training progresses, the generator is forced to keep closing the gap with real data, since the discriminator keeps getting sharper at spotting whatever gap remains.
GANs vs. diffusion models today
Diffusion models have largely overtaken GANs for high-quality image generation, mainly because GAN training is notoriously unstable — the two networks can fail to converge together — while diffusion training is comparatively steady and predictable, even though it requires more computation per image.
Frequently Asked Questions
What is a GAN?
A generative adversarial network — two neural networks trained against each other, a generator creating fake content and a discriminator trying to detect it, each improving because of the competition.
Are GANs still used for AI image generation?
Less so today — diffusion models have become the dominant approach for high-quality image generation because they train more stably, though GANs are still used in some specialized applications.