Is my image uploaded anywhere?

No. The AI model runs 100% in your browser using transformers.js — your image never leaves your device. Once the model has downloaded you can switch off your internet connection and it still works.

Why is there a download the first time?

The AI vision model (about 360 MB) is fetched once from this site and cached in your browser. Every generation after that is fast and offline — there is no per-image upload and no API call. Clearing your browser data will remove the cache and trigger a fresh download next time.

Which AI model does this use?

Microsoft's Florence-2, a modern vision-language model, run with transformers.js. By default it uses Florence-2-base-ft (about 360 MB); turn on "Higher accuracy" to switch to the larger Florence-2-large (about 1.1 GB) for richer, more accurate descriptions — the bigger model downloads only when you enable it. Either way it identifies subjects far more accurately than older captioning models and runs entirely in your browser (on your GPU via WebGPU, or your CPU as a fallback), with no server, no API call, and no account required.

What makes good alt text?

Describe the content and its purpose in one clear, specific sentence; don't start with "image of"; transcribe any important text shown in the image; and mark purely decorative images with an empty alt="" so screen readers skip them.

Do I need a powerful computer or a GPU?

No special hardware is required. When your browser supports WebGPU the model uses your GPU for speed; otherwise it falls back to your CPU, which still works but is slower. Either way it runs right inside the browser. Generation takes a few seconds on most modern devices; the Detailed option takes a little longer.

Can I do a whole folder of images at once?

This free tool handles one image at a time. Bulk alt-text generation and an accessibility audit for many images at once is the Pro flagship — see the Accessibility & Alt-Text Audit.

Accessibility

AI Alt-Text Generator

Drop in an image and get accurate, editable alt text generated entirely on your device — nothing is uploaded. Choose concise or detailed, then copy the text or the ready-to-paste <img> tag.

Runs entirely on your device — nothing is uploaded.

Drop your image here

or click to browse · JPG, PNG, WebP, GIF · Nothing is uploaded

The AI runs entirely in your browser. The first generation downloads the model (about 360 MB); after that it works offline.

⚠️ AI-generated alt text is a starting point, not a finished answer. It can miss context, misread complex images or infographics, and won't reliably transcribe text inside the image. Always read and edit it before publishing.

The AI Alt-Text Generator uses on-device AI to produce accurate, editable alt text for any image — no API key, no upload, no account — then outputs plain text or a ready-to-paste HTML img tag.

Key facts

On-device AI — image is never sent to any server or API
Choose concise or detailed output mode
Outputs plain text or a ready-to-paste img tag with the alt attribute
Supports JPEG, PNG, WebP, GIF and AVIF
100% in-browser — your image never leaves your device
Free, no account or sign-up required

Drop in an image

Choose or drag any JPG, PNG, WebP, GIF or AVIF image. It stays on your device — nothing is uploaded to a server. The AI vision model runs entirely inside your browser, so your images stay completely private.

Generate alt text on your device

The first time you generate, your browser downloads the AI vision model (about 360 MB) and caches it — after that it's fast and works offline. Pick Concise or Detailed and press Generate. Want richer descriptions? Turn on "Higher accuracy" to load the larger Florence-2 model (about 1.1 GB) — it only downloads if you enable it. Everything runs locally: it uses your GPU via WebGPU when available, and falls back to your CPU otherwise (a little slower). Generation takes a few seconds.

Review, edit and copy

The AI gives you a draft. Edit it so it's accurate and describes what matters in context, then copy the text or the ready-to-paste <img> tag. For purely decorative images, flip the Decorative toggle to get an empty alt="" instead.