AI Alt Text Generator

Alt text is the sentence a screen reader speaks aloud when it reaches your image – and the sentence a search engine reads instead of the pixels. Upload the image and a vision model writes it: a plain alt attribute, a longer description, or a caption to post with it. Run it free to see the wording; unlock the full text with credits.

What do you need written?

Leave it on Alt text for the HTML alt="" attribute – one sentence, no "image of". Pick Long description when the image carries information the surrounding text does not.

Drag & drop a file here, or

JPG PNG WebP Max 20MB • Processed on server • Deleted immediately

Written the Way Alt Text Should Be

One sentence, present tense, no "image of" or "picture of" – a screen reader already announces that it is an image. The model is told to state what is shown, plainly, so the result drops straight into the alt attribute.

Three Lengths, One Upload

Alt text for accessibility and SEO, a long description for images that carry information the page text does not, or a caption for social. Same image, different instruction to the model.

Private & Secure

Your image is sent to our AI server to be described. It is deleted immediately; the text is held only briefly so you can unlock it, then expires automatically. Nothing is shared.

How it works

What good alt text actually says

Alt text exists for the person who cannot see the image: a screen reader reaches the picture and reads your sentence out loud instead. That single job settles most of the questions people argue about. Do not begin with "image of" or "photo of" — the screen reader has already announced that it is an image, so those three words are a tax on every listener. Describe what is shown, in one plain sentence, in the order that matters: the subject first, then what it is doing, then only the context that changes the meaning. Keep it short — roughly a sentence, and under about 125 characters, which is where several screen readers break the text up anyway. Do not stuff it with keywords; search engines have read alt attributes for two decades and treat a keyword list as exactly what it is. Two special cases are worth knowing: an image that is purely decorative should have an empty alt (alt="") so the reader skips it entirely, and an image that carries real information — a chart, a diagram, a screenshot with text in it — needs more than a sentence, which is what the Long description mode here is for. The same sentence does double duty for search: it is the only text a crawler has for the image itself, which is why alt text and image metadata are the two things worth getting right before you publish.

Getting a usable sentence out of it

Pick the mode that matches the job, because the mode is the instruction the model is given. Alt text asks for one plain sentence for a screen reader. Long description asks for the subject, the setting, the colours and the action — use it for charts, diagrams, screenshots and product shots, where the image carries information the page text does not. Caption asks for the line a person would actually post with the photo. What a vision model sees is what is visible: it will describe two people at a table, but it does not know they are your co-founders, and it cannot read a name that is not in the frame. So treat the output as a first draft — correct the names, drop a detail that does not matter, and it is done in seconds instead of minutes. Where it earns its keep is volume: a product catalogue, an image-heavy archive, a blog with two hundred old posts whose images have no alt at all. Run each image, keep the sentence, move on. If you are working through a batch, the Image Compressor and the Image Resizer handle the other half of publishing an image well. Every run here is free to preview; credits only unlock the full text, and the pricing page shows what a credit costs.

Frequently Asked Questions

How do I generate alt text for an image?

Upload the image, choose what you need written — alt text, a long description, or a caption — and a vision model reads the picture and writes the sentence. You see a free preview of the wording straight away; unlocking the full text costs credits.

What makes alt text good?

One plain sentence that says what is shown, written for someone who cannot see the image. Do not start with "image of" or "photo of" — a screen reader already announces that it is an image. Keep it under about 125 characters, put the subject first, and do not stuff it with keywords.

Is the alt text generator free?

Running it is free and needs no account. The preview shows the opening of the text — enough to judge the wording, not enough to use. Unlocking the full sentence costs credits, which is what pays for the AI processing.

Does unlocking the text run the AI again?

No. The full text is generated on the first free run and held on the server under a one-time token. Unlocking reveals what is already there, so you are never charged for a second run of the same image.

What is the difference between alt text and a long description?

Alt text is a sentence, for images that illustrate. A long description covers the subject, the setting, the colours and what is happening — use it when the image carries information the page text does not, such as a chart, a diagram or a screenshot with text in it.

Can I use the output for SEO?

Yes. The alt attribute is the only text a search engine has for the image itself, so a clear, accurate description helps. What does not help is a list of keywords — that has been read as spam for years. Describe the image honestly and the SEO takes care of itself.

Will it get every detail right?

It describes what is visible, and it is good at that. It cannot know things that are not in the frame: it will say two people at a table, not that they are your co-founders. Treat the output as a first draft — check names and facts, then use it.

What AI model does it use?

Moondream 2, a small vision language model built to run efficiently and licensed under Apache 2.0, so its output is safe to use commercially. The mode you pick is the instruction it is given, which is why the three modes read so differently.

Are my images stored on the server?

Your image is sent to our AI server to be described and is not kept. The text is cached only briefly so you can unlock it, then it expires automatically. Nothing is shared or reused.

Scanly is an independent product. Moondream 2 is a third-party model we call through Replicate — an integration, not our brand. Scanly is not affiliated with, endorsed by or sponsored by Replicate or the model authors, and all model names and trademarks remain the property of their respective owners. AI Disclosure.