About this site
An alt text checker with somewhere to write. You supply the words — for an alt attribute, a caption or a long description — and this site reads them back the way assistive technology will, with the picture and its measurements sitting beside the field. It does not write descriptions of photographs for you: no captioning model runs here.
Why a whole site for one attribute
Because it is the piece of a web page most consistently left to whatever the camera called the file, and because the advice about it is scattered across specification documents, vendor help pages and blog posts that quote a character limit without saying where it came from. Somebody writing forty alt attributes on a Tuesday afternoon should not have to assemble that themselves.
The other reason is that the checking half is genuinely solvable. A great deal of what goes wrong in alt text is structural — a redundant opening, a file name, a keyword list, a missing pause — and structure is something a program can read reliably. What it cannot read is whether your sentence matches your photograph, and this site is careful about which side of that line it is standing on.
What it does, and the line it does not cross
It does this: reads a string the way a screen reader will and names every structural fault in it, with the edit that fixes it in the same sentence. Takes the measurements a page can honestly make from a picture without recognising anything in it — shape, tone, colour, light, layout, the camera’s own date and model — and, when you press for them, runs three small on-device detectors that place faces and people and estimate what is near or far. Assembles a draft from those facts and the subject you type, with every clause showing its evidence. Reads the printed English and Hindi text in a picture, on your device, with an open-source text recogniser. Loads three rulesets for three different jobs. Carries four written pages for the questions a findings list cannot settle.
It does not recognise what a photograph is of. That is the line, and it is worth stating in the same breath as the rest rather than in a footnote further down. Naming the subject — a dog, a harbour, a wedding — needs a vision-language captioning model; this site doesn’t run one yet, and it will not borrow one from a third-party model hub or send your picture to an API to get one. So the subject is the one thing you type, and the draft around it is built only from what was measured or detected.
None of which makes this a lesser tool than a captioner would be. A machine-written sentence still has to be read, corrected and judged by somebody who can see both the picture and the page it is on. That reading is the part this site is good at, and it is the part that decides whether the alt text was any use.
How it is made
A static export. There is no server behind this site, no route that accepts a request and no database anywhere. The picture is decoded by the browser you are already using, measured, and forgotten when you close the tab. That is not a policy decision that could be reversed later — there is no other architecture here to reverse it to.
Where else to go
- Alt textWrite the alt attribute yourself, with the picture open and the findings live.
- Alt text checkerPaste alt text you already have and read the findings. No picture required.
- WCAG 1.1.1 and GIGW, in plain termsWhat the success criterion asks for and what an Indian government audit checks.
- PrivacyWhat this site sees, and what it cannot.