Diffusion Model
Category: AI Architecture
📖 Definition
Diffusion models generate images by starting with random noise and gradually refining it into coherent images. The model learns to reverse a gradual noising process to create images from prompts.
🔑 Key Points
- Starts with random noise and denoises it step by step
- Training is more stable than GANs
- Powers Stable Diffusion, Midjourney, and DALL-E 3
- Can generate high-quality, diverse images
💡 Why It Matters
Diffusion models made AI image generation accessible and high-quality. Most modern image AI tools use this architecture.