MuhammadLab

Diffusion Models — Modern Generative AI (Explained)

An interactive simulation to demystify the core concepts of diffusion models.

The Diffusion Process

Diffusion models are a class of generative AI models that learn to create new data (like images) by understanding how to reverse a process of gradual data corruption.

Imagine taking a clear image and slowly adding random noise until it's completely obscured. Diffusion models learn to do the opposite: starting from pure noise, they iteratively remove noise to reveal a coherent image, often guided by a text prompt.

This simulation will guide you through these two core phases: **Noise Addition** and **Denoising**.

This canvas displays the image undergoing noise addition and denoising.

Original Image / Current State

Noise Addition

Denoising