Skip to content

Extract

Generate alt text for an image

Alt text is what a screen reader announces in place of an image, and what search engines read when they cannot see one. Writing it well is genuinely hard, and most images on the web have none. This tool drafts it for you using an AI vision model. It is one of only two ImageUtility tools that does not run on your device: it sends a small, reduced copy of your image to Anthropic's Claude API to generate the text, and it asks you before it does. Almost every other ImageUtility tool processes your image entirely in your browser.

Loading the alt text tool...

How to generate alt text for an image

  1. Drop an image onto the upload area, or choose a file from your device.
  2. Read the disclosure explaining what is sent and to whom.
  3. Choose whether to send the image, or cancel and keep it on your device.
  4. Review the suggested alt text, edit it if needed, and copy it.

Supported formats

This tool accepts JPG, PNG and WebP files. Everything is decoded by your own browser, so the formats available depend on what your browser can open - which for HEIC in particular varies between Safari and other browsers.

Frequently asked questions

Does this tool upload my image?

Yes, and it is the only ImageUtility tool that does. A reduced copy - no larger than 1024 pixels on its long edge, re-encoded and stripped of metadata - is sent to our server, which passes it to Anthropic's Claude API and returns text. Your browser never talks to Anthropic directly. No part of your image is sent until you press the button on the disclosure, and if you decline it there, your image is never transmitted at all. Cancelling after that button stops this page waiting for the answer; it cannot recall a copy that has already gone, and the page says so rather than claiming otherwise.

What happens to my image afterwards?

ImageUtility does not store it. It exists in server memory for the duration of one request, is never written to disk or to a database, and is never logged. Anthropic keeps it under their own policy: as of September 2026 their documentation states that commercial API inputs are deleted within 30 days, that retained data is not used for model training without express permission, and that content flagged by their automated trust and safety systems may be kept for up to two years. Those links are on the page. We describe their policy rather than promising deletion on their behalf.

Why can't this run in my browser like your other tools?

Describing an image in natural language needs a vision-language model, and the smallest ones useful for this are far too large to download to a browser. We looked at doing it locally and concluded it is not currently possible at acceptable quality, so we have made the trade explicit rather than pretending the tool is local.

Is the generated alt text good enough to use as-is?

Treat it as a first draft. It describes what is visibly in the image, which is the bulk of the work, but alt text depends on context: the same photograph needs different wording in a news article and a shopping page. Read it, edit it, and make sure it says why the image is there.

What if my image contains text telling the AI what to do?

It is ignored. Text inside an image is treated strictly as something to describe, never as an instruction to follow, and the response is validated against a narrow schema before you see it. This matters more than it sounds: screenshots are a common input, and an image is an easy place to hide an instruction.

Is there a limit on how many I can generate?

Yes. Calls to an external AI service cost money, so the tool is rate limited and the request size is capped. If you hit a limit the tool says so plainly rather than failing silently, and everything else on ImageUtility keeps working because none of it depends on this service.

Looking for something else? Browse all ImageUtility tools.