Skip to the main text
Foretaste Get Foretaste

Text in a picture

Extract text from an image without uploading it

Somebody sends you a scan. A receipt, a form, a photograph of a page. Your browser will show you the picture and nothing else, and every free way to get the words out of it starts with handing the picture to a website. Foretaste reads it on your machine instead, from a button, and the picture never moves.

Add to Chrome, free What you see for a picture

01

How it goes

  1. Open the picture with Foretaste.

    Right-click the image link and choose Open with Foretaste, or press the marker beside it. It opens in a panel over the page you were already reading.

  2. Press "Read the text".

    It sits under the picture with the rest of the facts about the file. The first press takes a few seconds to start the engine up.

  3. Take the words.

    Copy them, save them as a plain text file, or save them as JSON that records which picture and which language they came from. There is a search box over the result, which is the fastest way to answer "is the invoice number in here".

The text is shown in a monospace face on purpose. Recognition confuses l with 1 and O with 0, and a proportional font hides exactly the characters you are checking.

02

Nothing is uploaded, and for English nothing is downloaded either

The engine is Tesseract, compiled to WebAssembly, running in a worker inside your browser. It ships inside the extension: the worker, the engine and the English data are all files in the package.

So reading English text makes no network request at all. Not to us, not to Google, not to an OCR service. There is a test in the extension that fails the build if any path the engine is given for a built-in language points anywhere except at the extension itself, so this is checked rather than promised.

What Foretaste can and cannot see, in full.

03

It reads a picture you have not downloaded

This is the part the desktop OCR tools and the upload-a-file websites cannot do. They all need the picture on your disk first, which means downloading something you have not seen to find out what it says.

Foretaste starts one step earlier, so the same button works on:

The picture is fetched into the tab and read there. It reaches your disk only if you press Save, which is free and needs no site permission.

04

Fourteen languages, and what each one costs

English, French, German, Spanish, Italian, Portuguese, Dutch, Polish, Russian, Arabic, Hindi, Chinese (simplified), Japanese and Korean.

English is in the package. Everything else is a language pack of roughly one or two megabytes, fetched once from a button you press, and kept on your machine afterwards. It is never fetched because a picture happened to contain French, and the screen tells you the size before you press anything.

The list is short on purpose. A menu of a hundred languages is a worse way to find French than a menu of a dozen, and every entry here is a promise that the pack exists.

All of it is free, in every language. Pro is about digging into files, not about reading them.

05

Where it stops

It reads the whole picture, and it uses your processor. That is why it is a button rather than something that happens to every image you look at. A large scan takes seconds and a fair amount of memory.

Handwriting is not what this engine is for. It reads printed text. Neat handwriting sometimes comes out and usually does not.

Layout is not kept. You get the words, in reading order, not a table with its columns and not a form with its boxes.

A crooked or blurred photograph reads badly. Straighter and sharper is the whole trick. If nothing comes out at all, that is usually why.

It is a guess, and it says so. Recognition is not the same as a text layer being read out exactly. Check anything that matters, particularly digits.

If you are going to run OCR over a hundred files on your own disk, a dedicated desktop tool is the better answer, and this page is not going to pretend otherwise. Foretaste is for the one you have not downloaded yet.

A scanned PDF is the same problem

Most PDFs are not scans. Foretaste decides page by page: a page with a text layer is read out exactly, without drawing anything, and a page that turns out to be a photograph of paper is drawn and recognised. You are told which pages were which, because one of those is exact and the other is a guess.

Reading a PDF without downloading the whole file.

The questions people ask

Can I extract text from an image without uploading it anywhere?

Yes. Foretaste runs an OCR engine inside your browser, so the picture is read on your own machine and is never sent to a server. English is built into the extension, so reading English text makes no network request at all.

Can it read a picture I have not downloaded?

Yes, and that is the point of it. Right-click the image link and choose Open with Foretaste. The picture is fetched into the tab, read there, and never written to your disk unless you press Save.

Which languages can it read?

Fourteen: English, French, German, Spanish, Italian, Portuguese, Dutch, Polish, Russian, Arabic, Hindi, Chinese (simplified), Japanese and Korean. English ships in the extension. Any other language downloads a language pack of about one or two megabytes, once, from a button you press, and it is kept locally afterwards.

Does it work on a scanned PDF as well as an image?

Yes. Foretaste decides page by page: a page with a text layer is read out exactly, without any recognition, and a page that is a photograph of paper is drawn and recognised the same way an image is. It tells you which pages were which, because a text layer is exact and recognition is a guess.

Is reading text out of a picture free?

Yes, in every language, including the ones that need a language pack. Nothing about reading a file is behind the paid plan.