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