5.3. Image-Generation Solutions

💡 First Principle: Image generation reverses computer vision: instead of reading an image to produce information, a generative model reads a text prompt to produce a new image. The prompt is the specification — the more specific it is about subject, style, and composition, the closer the output matches your intent.

Why care? The syllabus explicitly separates "interpret visual input" (understanding, Section 5.1.2) from "create new visual outputs" (generation, here). The exam tests that you don't conflate them, and that you know image generation is prompt-driven creation.

⚠️ Common Misconception: Generating an image and analyzing an image use different model capabilities and flow in opposite directions. Generation: text in, image out. Analysis: image in, information out. A "vision-capable app" might do either or both, so read what the app is actually doing.