Drop images or a PDF here

or click to browse — several files at once is fine

Language

Add an image or PDF to start reading text

Extract Text from Images and Scanned PDFs

Drop in a screenshot, a photo of a page, or a scanned PDF, and the text inside it comes back as something you can select, copy and edit. Recognition happens in your browser — the file never leaves your device. Several images at once is fine, and a PDF is read page by page.

What This Tool Does

Reads printed text well

On clean screenshots and scans, recognition is close to exact. On photos with uneven lighting or a bit of perspective it holds up too, though a straight, well-lit shot always reads better than a crooked one.

Nine languages

English, Chinese (both simplified and traditional), Japanese, Korean, German, French, Spanish and Italian. Non-English options load English alongside, because real documents are full of order numbers, emails and amounts written in Latin characters.

Scanned PDFs, page by page

A PDF is opened and each page is read in turn, so a multi-page scan comes back as text without you exporting the pages yourself first.

Several files at once

Drop in a folder's worth of screenshots and they queue up. If one of them fails, the rest carry on — a single bad file does not stop the batch.

Nothing is uploaded

The recognition engine runs inside the page itself. Your images and PDFs are never sent anywhere, which matters when the document is an invoice, a contract or anything with personal details on it.

Copy or download

Take the result straight to the clipboard, or download it as a .txt file. With several files, each one is labelled so you can tell the results apart.

How to Extract Text from an Image

Add your files

Drag in images or a PDF, or click to browse. JPG, PNG, WebP, GIF, BMP and PDF all work, up to 50MB each.

Pick the language

Choose the language of the text in the picture — not the language of this page. Getting this right matters more than anything else for accuracy.

Copy or download the text

Results appear as each file finishes. Copy everything to the clipboard, or download it as a .txt file.

When This Is Useful

Text trapped in a screenshot

Someone sends a screenshot of an error message, an address or a code snippet. Rather than retyping it, pull the text out and paste it where you need it.

Scanned documents you cannot search

A scanned contract or receipt is just a picture as far as your computer is concerned. Reading the text out makes it something you can search and quote.

Photographed pages

A page of a book or a whiteboard after a meeting, photographed rather than typed. Get the words back without transcribing by hand.

Invoices and receipts

Pull out reference numbers and amounts without squinting at a photo and copying digit by digit.

Old paperwork being archived

When boxes of paper get scanned, the scans are images. Extracting the text is what makes the archive actually usable later.

Text in a language you are learning

Getting the characters out as text lets you look them up, which you cannot do while they are locked inside a picture.

Getting Better Results

Set the language first

This is the single biggest factor. Reading Chinese text with the English model set will produce nonsense, no matter how clear the image is.

Crop to just the text

Photos of a page usually include the desk, a hand, part of the room. Cropping down to the paper alone gives the engine less to be confused by.

Even lighting beats bright lighting

A shadow falling across half the page hurts more than the picture being a bit dark overall. Move so the light is even rather than strong.

Keep the lines horizontal

Text that runs at an angle is read noticeably worse. Straightening the photo before uploading is worth the few seconds.

Use the original file

A screenshot of a screenshot, or a photo forwarded through several chat apps, has been compressed each time. The original is always sharper.

Printed text, not handwriting

This works on printed and typed text. Handwriting is a genuinely different problem and the results will disappoint you.

Why Use This One

Your documents stay yours

Most online converters upload your file to their server. This one does not — the whole thing runs in the browser, so there is nothing to delete afterwards.

Free, no account, no watermark

No page limits, no signup wall halfway through, no upsell before you can see the result. The site is free and stays that way.

Works offline once loaded

After the engine has been fetched the first time, your browser keeps it. Later visits start immediately and work without a connection.

Batch and PDF included

Several images at once and multi-page PDFs are part of the free tool, not something behind a paid tier.

Chinese, Japanese and Korean actually work

CJK support is not an afterthought here. The spacing that recognition engines normally scatter between characters is cleaned up, so what you get is text you can use rather than text you have to fix.

Works on your phone

It all runs in a mobile browser, so you can read a photo straight from your camera roll.

Questions About Extracting Text

Yes. OCR — optical character recognition — is the general name for reading printed text out of a picture, and that is exactly what happens here. The difference from most OCR websites is where it runs: the engine is downloaded into your browser rather than your file being uploaded to someone's server.

On clean printed text — screenshots, good scans — it is close to exact. On our own test set, English printed text came back perfect and Chinese printed text was around 98% correct. Photographs with uneven lighting or slight perspective came in around 94–95%. Expect to skim the result rather than trust it blindly, especially for numbers.

No, and this is worth being clear about. The engine is built for printed and typed text. Handwriting recognition is a different problem needing different models, and trying it here will waste your time.

No. The text inside a table will be read, but the row and column structure is not preserved — you get the words, not a grid. The engine does not detect table structure at all, so there is no way to export usable Excel from it.

No. The recognition engine is downloaded to your browser and runs there. Your images and PDFs are read directly from your device and never sent anywhere. That is also why it keeps working if your connection drops mid-way.

The first time you use it, the browser downloads the recognition engine and the language data — about 6MB in total. After that your browser keeps it cached and subsequent runs start straight away. Each image then takes well under a second.

The language of the text in the image, not the language of this page. If you pick wrong the output will be nonsense even from a perfectly clear image. Non-English choices load English as well, so mixed documents with order numbers and email addresses still read correctly.

Yes. Each page is rendered and read in turn, and results appear page by page. Very long documents are capped at the first 100 pages, and you are told when that happens rather than being left to wonder.

You will see "No text found" rather than an error, because an image with no readable text is a normal result. Usually it means the picture is too blurry, the text is too small, or the wrong language is selected.

Because it should not have them. Recognition engines treat each Chinese character as a separate word and put spaces between all of them, which makes the result awkward to use. Those spaces are removed here — but spaces between Chinese and Latin text are kept, since those are real.

Yes. Drop in as many images as you like and they queue up, with results appearing as each finishes. If one file fails the rest continue, and you can stop the queue at any point without losing what has already been read.

Related Image Tools

Other tools for getting things out of images