Four answers, not one
A folder from a website is never one kind of picture. Product shots, a spacer, a scanned notice and a camera-roll export each want a different answer, and the page gives each its own: a sentence, an empty attribute, the picture’s own words, or a request for your subject. Every row prints the measurements that decided it — how much of the frame is on an edge, how many colours, how text-like it is — and how firm the call was.
Nothing is exported because it was guessed. Rows only reach the export when you approve them, and the CSV records for each one where its words came from and which decision it was given.
What it will not do
- It does not recognise what a photograph shows. Without a descriptive file name, a photo waits for your words.
- It does not upload anything, so it cannot fetch pictures from a website for you; open the files themselves.
- It writes embedded alt text into JPEG, PNG and WebP only.
Questions people ask
Are my pictures uploaded?
No. Every picture is decoded and measured in this tab, one at a time, and the exports are written in the tab. A folder of two hundred pictures never leaves your computer.
Where do the sentences come from?
From words a person already wrote, or from the picture's own text. A descriptive file name (mountain-lake-at-dusk.jpg) becomes a sentence and says it came from the name; a camera counter (IMG_4032.jpg) gets nothing. A picture that reads as a page of text can have its words read on this device. Nothing here recognises a subject, so every other row waits for yours.
What does “approved” mean?
Each row has a review state: untouched, suggested (the words came from a file name or the text reader), edited (you typed or changed something) and approved (you pressed Approve). The exports include approved rows only unless you untick that, and say how many were left out.
Can I mark several pictures decorative at once?
Yes. Tick the rows and press “Mark selected decorative”. Each row also has its own three answers — describe, decorative, inside a link — which go into the CSV and JSON as a mode column.
What can I export?
A CSV and a JSON manifest with the measurements behind every row, an HTML block of <img> tags with the alt text filled in (alt="" for decorative pictures), a ZIP of one .alt.txt per picture, and a ZIP of the pictures themselves with the approved alt text written into their IPTC accessibility fields (JPEG, PNG and WebP).
Is there a keyboard way through a long folder?
Yes. ↑ and ↓ move between rows, Enter opens a row with a larger preview and puts the cursor in its field, and Esc closes it again.
Where to take this next
- Product alt textProduct name, measured colour and view for each photo, exported as a store CSV.
- Embed alt textWrite alt text into a JPEG, PNG or WebP file's IPTC fields without touching the pixels.
- Alt text checkerPaste alt text you already have and read the findings. No picture required.
- When the right alt text is noneThe images that want an empty attribute, and why describing them is worse.