AI Image Alt Text Generator
A first draft of alt text for an image, written to the rules that matter: under 125 characters, no "image of", and DECORATIVE when the picture adds nothing.
This tool needs JavaScript: a model describes the image, because nothing else can.
A public URL. The image is fetched by the server, sent to the model once, and not stored.
Optional, and it changes the answer more than anything else: alt text describes what the image is doing on that page.
This is one of the few tools here that asks the server rather than your browser, because a browser cannot do it. Nothing you put in is stored; the request is rate limited by address and the answer is not cached.
Alt text describes what an image is doing on a page, which is why a model can only ever write the first draft: it can see what is in the picture and cannot know why the picture is there.
What it does well is the mechanical part, which is also the part people get wrong. Under 125 characters.
No “image of”. The text inside the image if that text carries meaning. And alt="" when the picture
adds nothing the surrounding words do not already say — which is the right answer far more often than
generators admit.
How to use
- Put in the address of the image.
- Add a sentence about what the page is about. This changes the answer more than anything else.
- Read the result before you use it.
Example
Image https://example.com/product/kettle.jpg
Type image/jpeg, 184 KB
Alt text Stainless steel kettle with a wooden handle on a kitchen worktop
length 63 characters
To paste
html <img src="…" alt="Stainless steel kettle with a wooden handle on a kitchen worktop" />With “the page is about the handle being heat-resistant” as context, the same image gets alt text about the handle instead. That is the difference between description and alt text.
Pitfalls
Alt text is contextual, and a model cannot see the context. The same photograph on a product page, a blog post about materials and a news article about a factory needs three different sentences. Give the context field something to work with.
alt="" is a real answer. A decorative image with an empty alt is skipped by a screen reader, which
is correct and quieter than hearing a description of a divider. This tool says DECORATIVE when it thinks
so; take it seriously.
Never start with “image of”. A screen reader already announces it as an image, so the words are wasted twice.
Text in an image has to be in the alt text. A poster, a chart label, a button rendered as a picture: if the words matter, they belong in the attribute.
A chart needs more than alt text. The short version goes in the attribute and the numbers go in the page, in a table or a paragraph. A screen reader user should not be the only person who cannot read the data.
This spends the site owner’s money. Each call is a model request with a cost, capped daily and monthly. The receipt at the bottom of the output says exactly what it cost.
Nothing is stored. The image is fetched, sent once, and forgotten. If the image is private, the same caution applies as with any tool that fetches a URL: the server requests it, so it has to be publicly reachable.
Compatibility
The image is fetched by the server — the same guard as the other server tools: http and https on the standard ports, private and reserved addresses refused, redirects followed by hand — checked to be an actual image, capped at 4MB, and sent to the model once.
The rules in the prompt are printed with the result, because a tool that puts words in your page should show what it asked for. The model is told to answer with the alt text alone, which is why the output is the sentence rather than a paragraph about the sentence.
Cost is counted from the tokens the provider reports, charged against a monthly ceiling the site owner sets. When the ceiling is reached the tool says so and stops; every other tool on this site runs in your browser and is unaffected.
The key is held in wp-config.php or in an option, and never leaves the server: there is no endpoint
that returns it and no field that renders it.
Frequently asked questions
Is the alt text good enough to use as it is?
Why 125 characters?
figcaption, where everybody can read them.