Diffusion Model
A diffusion model is a generative AI model that creates images, audio or video by learning to reverse a process that gradually adds random noise to data.
What is Diffusion Model?
Diffusion models power most leading image and video generators. They produce high-quality, varied outputs but need many computational steps per sample, which newer techniques reduce.
How it works
During training, noise is added to examples step by step and a network learns to remove it; to generate, the model starts from pure noise and denoises it, guided by a text prompt.
← All Artificial Intelligence & Superintelligence definitions · ← Back to the A–Z definitions