Les registres d'une image qui explique
A generated image that explains something is not an illustration: it belongs to a **register**, and each register answers a different question. Five recur, and naming them is enough to get the right image. - **The exploded view** separates the object into parts laid out in assembly order. It answers "what is it made of". - **The cutaway** slices the object and shows the inside in place. It answers "how does it fit together inside". - **The x-ray** keeps the silhouette and reveals the contents through transparency. It answers "what is behind the surface". - **The technical drawing** converts the object into a dimensioned outline on a plain ground. It answers "what are the sizes and proportions". - **The anatomical plate** labels each part with its name. It answers "what is this called". The nuance that matters: exploded view and cutaway look alike but do not serve the same intent. The exploded view shows the **inventory**, the cutaway shows the **relationship between parts**. Asking for one when you wanted the other gives you a pretty, mute image.
Strengths
- Five names are enough, and they are technical illustration terms every image model already knows
- The register is chosen from the question asked, not from the subject depicted
- Separates an informing image from a decorating one, before spending a generation
Limitations
- None of these registers suits a process over time, which calls for a vector diagram, not an image
- The labels on a generated anatomical plate are often wrong and must be checked one by one
Best for
- Illustrating an article or training material with an image carrying verifiable information
- Deciding upstream whether the need calls for a generated image or a hand-drawn diagram