Diffusion Model

A diffusion model is a generative model trained by taking real data, adding noise to it in many small steps until only noise remains, and teaching a neural network to reverse each step. To generate something new, it starts from random noise and repeatedly removes noise, guided by a text prompt or other conditioning, until a clean image, audio clip or video emerges.

Why it matters

Diffusion models power many modern image and video generators. They train more stably than GANs and cover varied outputs well, but sampling needs many network evaluations, so generation is slower and costlier than a single forward pass. Techniques such as latent diffusion work in a compressed space and fewer-step samplers to cut that cost.

Example

Training:  clean image -> + noise (step 1) -> ... -> pure noise (step T)
           network learns to predict the noise added at each step
Generating: pure noise -> denoise (T steps) -> image matching the prompt

Common mistakes

• Assuming the model stores and pastes training images. It learns statistical patterns, though memorisation can occur in some cases.
• Expecting precise text rendering or counts, which are common weaknesses.
• Ignoring licensing and consent questions around training data.

Related

• GAN
• Generative AI
• Multimodal AI
• Neural network

This page is general information written by our editorial team and updated when documentation changes. Servers and providers differ, so test changes safely first. See our editorial policy and disclaimer.