ITG NowHome › AI glossary › Diffusion model
What is a diffusion model?
Short answer
A diffusion model is an AI that makes images by starting from random noise and cleaning it up step by step.
- Category Models
- Also called diffusion
- Updated 20 Oct 2026
Diffusion model, explained simply
During training, the AI learns how a picture turns into noise. Then it learns to run that in reverse.
To make a new image, it starts with static, like an old TV, and slowly sharpens it into a picture that matches your words.
Many popular image and video makers use diffusion.
On ITG Now, our free tool guides cover image and video tools that use diffusion models.
Related terms
Text-to-imageText-to-videoInpaintingGenerative AIAI model
Questions people ask about Diffusion model
Is a diffusion model the same as an LLM?
No. LLMs mostly work with text. Diffusion models are best known for images and video.
Why is it called diffusion?
The name comes from physics, where things spread out. The AI learns to reverse that spreading.