🔮The Codex
Diffusion Model
The AI technique behind most modern image and video generation tools.
📖 Apprentice Explanation
Diffusion models create images by starting with random noise and gradually refining it into a clear picture. It's like sculpting — starting with a rough block and carving out details.
🧙 Archmage Notes
Diffusion models learn to reverse a noise-adding process through denoising score matching. Key architectures include DDPM, latent diffusion (Stable Diffusion), and consistency models. Guidance techniques (classifier-free) control generation quality.