📖 Step 9: AI/LLM#320 / 350

Diffusion Model

Diffusion Model

📖One-line summary

A generative model that creates images by progressively removing noise.

💡Easy explanation

A generative model that starts from pure noise (TV static) and gradually refines it into a clean image. Stable Diffusion, DALL-E, and Midjourney all work this way.

Example

노이즈에서 이미지로

📺

노이즈

🌫️

...

🌤️

...

🖼️

완성

단계마다 노이즈 제거 → 이미지 등장