Generate SEO-ready alt text for your images with AI — free, in your browser. Nothing is uploaded: the captioning model runs on your own device. Drop up to 10 images at once and export the results as CSV.

Drop images here or click to choose (up to 10)

JPG, PNG or WebP · processed locally · nothing uploaded

First use downloads the AI model once (cached afterwards). Chrome or Edge with a modern GPU gives the best captions (Florence-2 model); other browsers automatically fall back to a lighter model.

Why alt text matters for SEO and accessibility

Alt text is one of the few places where accessibility and SEO point in exactly the same direction. Screen readers speak it aloud to visually-impaired visitors, and search engines use it to understand what your images show — which drives image-search traffic and strengthens the topical relevance of the page. Yet most sites either leave alt attributes empty or stuff them with keywords, and both hurt.

This tool generates natural, descriptive alt text using a vision-language model that runs entirely in your browser. Because nothing is uploaded, you can safely run client images, unpublished product photos or private screenshots through it.

How to write good alt text (and how this tool helps)

  1. Describe what is actually in the image — the generated caption gives you a factual base to start from.
  2. Keep it under ~125 characters. The tool trims captions to this length automatically, because screen readers may truncate longer text.
  3. Skip “image of” / “photo of”. Screen readers already announce it as an image; the tool strips these prefixes for you.
  4. Add context the AI cannot know. If the photo shows your product or a specific person, edit the caption to name them.
  5. Copy the ready-made HTML snippet or export everything as CSV for bulk upload to your CMS.

Frequently asked questions

Are my images uploaded to a server?

No. The captioning model downloads to your browser once and runs locally. Your images never leave your device.

Which AI model is used?

On browsers with WebGPU (Chrome, Edge) it uses Microsoft’s Florence-2, an MIT-licensed vision-language model. Other browsers fall back to a lighter ViT-GPT2 captioning model.

Is there a limit on how many images I can process?

You can process 10 images per batch and run as many batches as you like. There is no daily cap because there is no server cost.

Should alt text include keywords?

Only when they describe the image naturally. Keyword stuffing in alt attributes can hurt rather than help. Describe the image first; if your target keyword genuinely belongs, include it once.