SantageAI Glossary › Diffusion Model
AI Glossary

What is a Diffusion Model?

A diffusion model is a type of generative AI that creates images, audio, or video by starting from random noise and gradually refining it into a coherent result.

What is the core idea behind diffusion models?

A diffusion model learns to reverse noise, turning static into a picture step by step.

How do diffusion models differ from related concepts?

ConceptDifference
Diffusion vs GANsGANs generate in one pass using two competing networks. Diffusion models refine over many steps and are more stable to train.
Diffusion vs TransformersTransformers dominate text. Diffusion models dominate image and video generation, though the two are increasingly combined.
Diffusion Model vs LLMLarge language models predict the next token of text. Diffusion models denoise pixels or audio into media.

How do diffusion models work?

What do diffusion models create?

Why are diffusion models important?

Diffusion models are the technology behind the AI image and video boom. They produce higher quality and more diverse results than earlier methods, and they underpin a fast-growing market of creative tools.

How are diffusion models used in practice?

They are used for image and video generation, editing and inpainting, design, and scientific applications like molecule and protein design. Limitations include compute cost, slow multi-step generation, and copyright and safety concerns around training data.

Frequently Asked Questions

Are diffusion models better than GANs?
For most image generation today, yes. Diffusion models are more stable to train and produce more diverse, higher-quality results, which is why they have largely replaced GANs.
What is Stable Diffusion?
Stable Diffusion is a well-known open diffusion model for generating images from text prompts, notable for being freely downloadable and runnable on consumer hardware.