GANs

← Back to the article

GANs use two AIs: one makes fake examples, and the other tries to spot the fake.

What it does

GANs use two AIs: one makes fake examples, and the other tries to spot the fake.

Generator: The creator • Discriminator: The judge

How it works

  • The generator makes a fake item.
  • The discriminator judges whether it looks real.
  • Both improve by competing.

Uses and applications

  • A game studio can make sample textures.
  • A design team can create example faces or objects for testing.
  • A research team can make practice images when real examples are limited.

In short

Picture a creator AI and a judge AI training together.

Related methods

  • CycleGAN: Uses GANs to translate between two image styles.
  • Diffusion: Generates by removing noise step by step instead.
  • Deepfakes: A common, risky application of generative models.

Back to Image and content generation · All topics

The interactive demonstration and the Python walkthrough for this topic need JavaScript.