Caption vs Alt Text: Two Different Jobs

task-specific prompting

Part of: Caption & Alt-Text Generator

A caption and alt text describe the same picture, but they serve different readers doing different jobs. That means two prompts, not one prompt reused twice. Two jobs, one image - A caption is for a sighted reader scanning a page. It can be vivid and a little editorial, there to add context or personality ("A golden retriever bounds through morning fog."). - Alt text is for a screen reader user who cannot see the image. It has to replace the picture: plain, complete, and short enough to read aloud without wearing out its welcome ("A dog running through a foggy field."). Send the same "describe this image" prompt for both and you get one mediocre result trying to do two jobs. The fix is two separate instructions, picked by mode. Writing task-specific prompts Calling with the right prompt Why "do not start with 'image of'" is in the prompt Screen reader software already announces "image" before it reads alt text. Alt text that starts with "image of" doubles that up and wastes the listener's time. A sighted person would never think to ask for this rule. It comes from knowing the audience, not the picture. The mental model to keep The picture doesn't change between the two calls. The a

Challenge: Choose the Right Prompt