What changes when you draw a box
The studio is the same one the alt-text page opens, with the region tool already switched on. As soon as a box is drawn, every measurement is taken again inside it: the palette ribbon shows that area’s colours, the light and contrast are that area’s, only the rows of text inside the box are counted, and the face and person finders report what falls inside it. The draft sentence says it is about a selected region, and the region’s size is shown in pixels.
You still name the subject. That is not a missing feature dressed up: a page that guessed who the person in a group photo was, or what a map shows, would be wrong often enough to be worse than a blank. The measurements are the part a machine can get right, and they are the part people otherwise have to squint for.
What it will not do
- It does not recognise or name what is inside the box — you do.
- It never identifies a person, or guesses age, gender or emotion.
- It does not crop or change the picture; the box only decides what is measured.
Questions people ask
Does it say what is in the box?
No. Nothing on this site recognises what a picture shows, and this page is honest about that: you name the subject. What it does is measure the area you drew — its colours by name, how light and how contrasty it is, where its rows of text are, and, if you run the small detectors, how many faces and how much of a person are inside it — and build the rest of the sentence around your words.
When would I describe only part of a photo?
When the page is about one part of it: one person in a group photo (never named by the page), a section of a map, the detail of a product, the one row of a screenshot that matters. Alt text for a cropped detail should describe that detail, and the measurements should come from it, not from the whole frame.
How do I draw and adjust the box?
Drag across the picture. Drag the corners to resize it, or use the arrow keys (Shift for bigger steps, Alt to resize) and Delete to clear it. Every change is a step you can undo. The R key turns the region tool on and off.
Is the photo uploaded?
No. The picture is decoded and measured in this tab. If you run the detectors, their model files are downloaded to your browser once and run there; the photo still goes nowhere.
Where to take this next
- Alt textWrite the alt attribute yourself, with the picture open and the findings live.
- Long descriptionThe paragraph a chart or a diagram needs when one attribute cannot hold it.
- Measured image tagsBuild honest tags from colour, tone, shape, transparency and detail — with no subject guessing.
- Where an image-description model guessesRecognise uncertain objects, relationships and context in machine-written photo descriptions.
