This alt text generator writes image descriptions for websites, shops and documents in bulk. Drop a folder of images, download the captioning model once, and get a short description for every picture that you can edit before exporting as CSV, JSON or ready-made img tags. The built-in draft mode works with no download. Images never leave your browser.
| Weak | Better |
|---|---|
| image.jpg | Red bicycle leaning on a brick wall. |
| Picture of a dog | Golden retriever catching a frisbee in a park. |
| Chart | Bar chart: sales doubled from 2022 to 2024. |
One concise sentence about what matters in the image for the reader, without “image of”; decorative images get an empty alt.
Usually close, but always review it: models can misread scenes and don't know the image's purpose on your page.
Around 125 characters or less works well with screen readers.
No. Files and text are processed in your browser. Only the model itself is downloaded, once, when you ask for it.
Yes. The Downloaded models panel lists what this tool stored and deletes it after a warning; you'd need to download it again to use it.
Images are captioned by a ViT-GPT2 vision-encoder-decoder model through Transformers.js in a Web Worker; the caption is tidied into alt text (leading “a picture of” removed, capitalised, ended with a full stop and trimmed to the chosen length at a word boundary). The built-in method needs no model: it uses words from a descriptive file name, the image's shape and its dominant named colours from a sampled palette. Every result is editable, can be marked decorative (empty alt) and exported as CSV, JSON or HTML img tags.