Skip to content
BENIBIX Learn

Diffusion Model

AI📈 Machine LearningNeural Networks

Also known as Denoising Diffusion Model

It starts with random dots. Step by step, it cleans them up into a picture.

Think of it like

Think of it like a fuzzy picture. It gets clear, one step at a time.

Example

Lucy types 'a red fox reading a book'. A diffusion model starts with static and, over 50 steps, turns it into that picture.

Why it matters

Most of today's image and video makers are built on diffusion models. They took over from GANs.

Definition

A diffusion model makes images by removing noise from them, step by step. In training, it takes real images, adds random noise to them, and learns to undo that noise. To make a new image, it starts from pure noise. Then it cleans it up over many steps until a picture appears. A text prompt can steer which picture it makes.
Each square has a little less noise than the one before. The last square is the finished picture.

Common mistake

Expecting a clean picture after a few steps. Quality comes from many small steps, so fewer steps is faster but rougher.

Key takeaways

  1. 1
    Learns to remove noise. From noisy real images.
  2. 2
    Starts from static. Pure random noise.
  3. 3
    Cleans step by step. Until a picture appears.

Compare

Learn it properly