MotionShine01 / Drop
Add your pictures
Drop many files or a whole folder, or paste a screenshot. Each one gets its own card.
Free tool
Alt text generator, for a whole folder of images.
Drop pictures or a folder and get a first draft of alt text for each one. Choose short alt text, a fuller description, or an empty alt for pictures that only decorate. Set a length cap, fix any line in its own box, then copy ready-made img tags or save a CSV with the file name and the alt text. Microsoft's open Florence-2 model looks at the pictures in this tab, so nothing is uploaded. No sign-up, no limit.
See the HTML
Settings
The cap is a guide. Text is cut at the end of a sentence or at a word. If you type more than the cap, the counter turns amber.
This runs in your browser. Your pictures are never uploaded. A model can describe what a picture shows, but it does not know the page it will sit on. Read every line yourself.
How it works
Three steps, nothing to install.
02 / Check
Read and fix each line
Pick short, descriptive or decorative, set the cap, and correct the text in each box. Tick Decorative for pictures that only decorate.
03 / Copy
Take it with you
Copy one line, copy all the img tags, or download a CSV with the file name and the alt text.
Guide
How to write good alt text, and where an AI draft helps.
An alt text generator looks at an image and writes a short text alternative for it, the words a screen reader speaks and a browser shows when the picture does not load. This one writes a first draft for each picture, in bulk, and leaves the final decision to you. That matters, because the right alt text depends on where the picture sits and why it is there, and a model only sees the picture.
How to use the alt text generator
- Drop pictures or a folder on the box, click it to choose files, press Choose a folder, or paste a screenshot with Cmd+V or Ctrl+V.
- Wait for the AI model to download the first time. Your browser keeps it, so the next visit starts at once.
- Pick a Style: Short alt text, Descriptive, or Decorative, empty alt. Changing it rewrites every line you have not edited. A line you typed in is never overwritten.
- Set the Cap, in characters, or leave it at the starting value.
- Read each box. Correct names, numbers and anything the picture means in your page. Tick Decorative on a card if the picture adds nothing a reader needs.
- Press Download CSV for a file with the columns filename and alt, or Copy all as HTML for one img tag per picture.
What the three styles do
| Style | What you get | Use it for |
|---|---|---|
| Short alt text | The model's one line caption, with openings like "The image shows" cut | Most photos in articles, product lists and galleries |
| Descriptive | The model's longer description, cut to the cap | Pictures where the detail is the point, and as a starting text for a longer description |
| Decorative, empty alt | An empty alt attribute, and the model is not run | Dividers, backgrounds and pictures whose meaning is already in the text next to them |
What the standards say about length
Neither of the two main guides gives a character limit. The W3C Web Accessibility Initiative asks for "at least a short description conveying the essential information" for an informative image, and for an empty alt when a picture only decorates (W3C WAI, Images tutorial, page updated 8 Apr 2026, read 3 Nov 2026). WebAIM says that "typically, only a few words are necessary", and that you should leave out phrases such as "image of" because screen readers already announce an image (WebAIM, Alternative Text, page updated 19 Oct 2021, read 3 Nov 2026).
The figure of about 125 characters that many tools quote is a rule of thumb, not a requirement. That is why the cap here is a number you can change. A long cap with a long description is fine when the picture carries real information, for example a chart, but then a short alt plus the data in the page text is usually better.
When to use an empty alt
Both guides say a picture that is only decoration should have an empty alt, written alt="", so a screen reader skips it. The same photo can need real alt text on one page and an empty alt on another. WebAIM gives the example of a portrait that is named in the caption next to it. A picture that is the only content of a link or a button is the opposite case: it needs text that says where the link goes, and an empty alt is wrong there.
Pictures that contain text
W3C WAI says the text alternative of an image of text should contain the same words as the picture. Tick Also look for printed text and the page will read any words in each picture with the same text reader as the image to text tool. Words it is sure about appear under the card as a suggestion. Add to the alt text puts them after the description. Use as the alt text replaces the description. Check them against the picture, because text reading makes mistakes on small, curved or stylised lettering.
A worked example
A photo of a café counter on a contact page. A short draft might read "A coffee machine and cups on a wooden counter." On a page about opening hours, that is probably too much and the picture may be decorative. On a page reviewing the café, "Espresso machine on the wooden counter at the Mill Street café" tells more. The tool cannot know which page you are on or the café's name. You add it.
What the model cannot do
- It does not know who people are, and does not try to name them.
- It can get small things wrong: counts, colours, left and right, and unusual objects.
- It does not read charts, maps or diagrams as data. Describe the finding yourself.
- It writes English only.
- It cannot tell whether a picture matters on your page. Only you can.
For a plain description to reuse as a prompt, try image to prompt. To get the words already printed in a picture, use image to text.
Questions
Good to know.
Which AI model writes the alt text?
Florence-2 base, an open vision model by Microsoft released under the MIT licence, in the version converted for browsers by the onnx-community project. This page uses its caption and detailed caption tasks.
What does the first use download?
The model, from Hugging Face, once. It is about 360 MB with a graphics card (WebGPU) and about 275 MB without. Your browser keeps it, so later visits do not download it again. If you tick Also look for printed text, the text reader and its English data are loaded from this site too.
Are my pictures uploaded?
No. The model runs in this tab. The pictures, the alt text and the file names stay on your computer. Only your settings (style, cap, the two ticks) are remembered in this browser.
Can I trust the alt text it writes?
Treat it as a draft. It describes what is visible, not why the picture is on your page, and it can be wrong about small details. Read every line, and fix names, numbers and anything that matters in context.
How long should alt text be?
There is no official character limit in the W3C WAI or WebAIM guides read on 3 Nov 2026. Both ask for short, useful text. About 125 characters is a common rule of thumb, so it is the starting cap here, and you can change it.
Can I do a whole folder?
Yes. Drop the folder on the box or press Choose a folder. Pictures in sub-folders are included, in name order. They are done one after another, and you can press Stop at any time.
What does Decorative do?
It sets an empty alt attribute for that picture, so screen readers skip it. Tick it on a card, or choose Decorative as the Style to do it for every picture.
What is in the CSV and the HTML?
The CSV has two columns, filename and alt, with the file name as it was added (a folder path for folder pictures). The HTML is one img tag per picture with src set to that file name, the alt text, and the width and height if that box is ticked. Change the src to the real address on your site.
Does it work for languages other than English?
The descriptions are in English only. You can type any language into the box. Printed text is read in English.
