AI alt text and picture generation in Publish by Zaweba

Pictures carry most of a poster's message, yet for anyone using a screen reader they say nothing unless they have a description. Publish by Zaweba helps on both sides of that with its optional AI picture tools: one writes descriptions for pictures you already have, the other makes new pictures for empty frames.

Like every AI feature in Publish, these need your own AI account or a model on your PC, and they're greyed out until you connect one.

Describing pictures for alt text

Every picture in Publish has an optional alt text property, shown and editable in the Properties panel. Alt text is a short description of what a picture shows, for people who can't see it. Writing it for every photo in a newsletter is a worthwhile but slow job, and Describe pictures speeds it up.

Select the pictures you want described, or select none and it works on the pictures on the page. Publish sends them as small JPEGs to a model that can see images, and gets back a description of each. Use as alt text then stores one description per picture in its alt text property. Because only small copies are sent, your full-size pictures stay on your PC.

Read each description before keeping it. A model can describe what's in a photo but not who the people are or why the picture matters. A description such as a group of people holding a cup in front of a clubhouse is accurate but plain; adding that it's the under-twelves team with the league trophy makes it useful. Keep alt text short and focused on what the picture is there to say.

Generating a picture

Generate picture creates a new image from a description and puts it into the page as a PNG. If an empty picture frame is selected, it goes into that frame; otherwise Publish makes a new frame for it. That's handy for a background texture, a simple illustration for a children's party invitation, or a decorative image for a menu when you don't have a suitable photo.

If you're not sure what to ask for, Suggest a picture prompt reads the text on the page and writes a prompt for you, which you can use as it is or adjust. It's a good way to get a picture that matches the theme of the words already there.

Using generated pictures well

Generated pictures suit decoration and mood better than facts. For a real event, a real place or real people, use real photographs. A made-up picture of your village hall won't look like your village hall, and readers notice. Use generation for backgrounds, borders, simple illustrations and general scenes, and keep real photos for anything that has to be true.

Check the result closely at full size before printing, since generated images can include odd details. And remember that once a picture is in a frame, the normal picture tools apply. You can crop it inside the frame, set its wrap, adjust brightness and contrast or recolour it, or send it to Paint by Zaweba for editing. Every AI change is a single undo step if you'd rather not keep it.

Where these tools fit

Describe pictures is about making a publication readable by more people; Generate picture is about filling gaps in the design. Both work best as helpers inside a layout you control. Choose a model that can handle images for these actions, since not every model can, and check the Context sent in the Ask AI panel to see what's being sent. Used that way, they add a little extra polish to Publish by Zaweba without taking any decisions out of your hands.