During training, the model learns how to remove noise from images. To generate something new, it starts with pure static and removes noise step by step, guided by your prompt.
Example
Many popular image and video generators are built on diffusion models.
Why it matters to you
It explains why the same prompt can give different results each time, and why “steps” and “seed” settings exist.
Seen in these tutorials
Related terms
SeedA number that sets the starting randomness of an AI generation./knowledge-base/seed/Text-to-ImageCreating an image from a written description./knowledge-base/text-to-image/Text-to-VideoCreating a video clip from a written description./knowledge-base/text-to-video/Generative AIAI that creates new content: text, images, video, audio or code./knowledge-base/generative-ai/
