Diffusion model: A diffusion model is a type of generative AI that creates images, audio, or video by starting with random noise and removing it step by step until a coherent result matching the prompt emerges. Most image generators of recent years are built this way.
During training, the model watches real images get gradually destroyed by noise and learns to reverse each step. At generation time it runs that process backward: begin with a canvas of static, and over a few dozen steps, nudge it toward something that looks like 'a lighthouse at dusk, watercolor.' A text encoder connects your words to the visual concepts.
Stable Diffusion, Adobe Firefly, and most video generators are diffusion models, and the approach also shows up in audio and music generation. Some newer image systems use different techniques, so the term is not universal, but it is the workhorse of the field.
Diffusion explains a few quirks you will notice: hands and text used to come out wrong because they are fine detail the denoising process handles last; small prompt changes can produce very different images because the starting noise is random; and generating takes more compute per image than a text reply. Many tools let you fix the random seed to get repeatable output.
Example at work
A product designer uses a diffusion-based tool to explore packaging concepts. She generates twenty variations of a box design from one prompt, picks three directions, and then refines each with follow-up edits. The concepts go into a review deck clearly labeled as AI-generated explorations, not final artwork.
Why it matters
You do not need to know the math, but knowing that images come from denoising randomness explains why results vary, why iteration is normal, and why a good prompt describes the whole scene. That knowledge makes you faster at getting the image you actually want.
Related terms
- Text-to-imageText-to-image is generative AI that creates a picture from a written description. Tools such as Midjourney, Adobe Firefly, and the image generator built into ChatGPT turn a prompt like 'a product photo of a blue water bottle on white' into a finished image.
- Generative AIGenerative AI is a class of AI models that create new content, including text, images, code, audio, and video, in response to a prompt. Chat assistants like ChatGPT, Claude, and Gemini and image tools like Midjourney are generative AI.
- Neural networkA neural network is a computing system made of layers of simple connected units that pass numbers to each other, with connection strengths (weights) adjusted during training until the network produces useful output. It is the building block of modern AI.
- Training dataTraining data is the collection of examples an AI model learns from, such as web pages, books, code, images, or labeled records. What is in that data, and what is missing, shapes everything the model knows, how it writes, and which mistakes it makes.