Answer
The model is trained to reverse a noise adding process step by step, learning how to remove noise gradually until a coherent image or output emerges.
Real-world example Diffusion models power many modern AI art tools that turn a text description into a detailed, realistic image.