← Library · Advanced concept
Diffusion Models
Diffusion models are a class of generative models that learn to reverse a gradual 'noise' process. They start with random noise and progressively refine it, guided by a learned denoiser neural network, to generate coherent and diverse data like images, audio, or video, often achieving very high fidelity.
In plain terms
Imagine starting with static on a TV screen and gradually clearing it up, learning what each blurry pixel should be to form a clear image.
Why it matters
They excel at generating high-quality, diverse content, opening new frontiers for creative AI applications and robust data augmentation.
Learn one new AI thing every day.
Daily Deck sends you seven plain-English cards like this every morning. Free.
Start free