Saved

Press the star next to a tool's name to save it.

All my tools

Image to Text (OCR)

Use this image to text OCR tool to read the words in a photo, screenshot or scan and copy them as text. The Tesseract engine runs in your browser, so your picture is never uploaded. English only.

Everything happens in your browser. Your files and pictures never leave your device.

English text only. The first picture takes a few seconds longer while the reader loads.

Text found in the picture

Add a picture to read its text.

Image to text OCR reads the words in a picture and gives them back as text you can copy, edit and search. This page runs Tesseract, a well-known open-source OCR engine, inside your browser. You can use a photo of a page, a screenshot, a scan or a sign, and the picture is never uploaded, because the engine and its English language data are downloaded from this site and run on your device.

Optical character recognition turns images of typed, printed or handwritten text into machine-readable text. [1] It saves retyping when you need the words from a receipt, a slide, a book page or a screenshot of an error message.

How this image to text OCR works

Tesseract was developed by Hewlett-Packard, released as open source in 2005 and sponsored by Google from 2006. Version 4 added a recognition engine based on LSTM neural networks. [2] Tesseract.js is a WebAssembly version of it that runs in the browser. [3] This tool uses the English "fast" model, a compact version of the language data that trades a little accuracy for speed. [4]

When you add a picture, the engine loads once, reads the picture and returns the text with a confidence score. The words appear in a box. You can copy them or download a .txt file.

Step by step

  1. Choose a picture, drop it on the box, or paste a screenshot.
  2. Wait while the reader loads. This takes a few seconds the first time, and a progress line shows the percentage.
  3. Check the text in the box and fix any mistakes.
  4. Copy the text or download it as a .txt file.

Getting accurate results

Results depend heavily on the picture. OCR works best on typed or printed text, and handwriting, especially joined-up writing, is read much less well. [1] Tesseract also benefits from images where the text is large enough, upright, evenly lit and free of dark borders. [2] Common steps in OCR software include straightening a tilted page and turning the picture black and white. [1] You can do the same before you upload: crop to the text, rotate the picture upright and avoid shadows.

PictureExpected result
Screenshot of typed textUsually good
Clear photo of a printed pageGood, if straight and well lit
Small or blurry textPoor; use a larger picture
HandwritingOften poor
Photos with curved pages or glareMixed
Text in other languagesNot supported here; English only

Language and layout

This version reads English only, which keeps the download small. Letters with accents may be missed or replaced by plain ones. The text is returned line by line. Columns, tables and text on busy backgrounds may come out in an odd order, so check them against the picture.

Privacy and speed

The reader, the engine code and the English data all come from this website, and the recognizing happens on your device. No picture and no text is sent anywhere. The files are about 8 MB in total on the first use. After that the browser can keep them for later visits. Large pictures take longer, especially on a phone, so crop to the part you need.

What the confidence score means

Along with the text, the engine reports how sure it is, as a percentage averaged over the whole picture. A high score means most characters were recognized clearly, and a low score means the picture was hard to read. The score is a guide only. A page can score high and still contain a wrong digit, so the check against the picture is always worth doing. If the score is low, try a larger picture, straighten it, crop away the background and try again.

What you can use the text for

  • Copy a paragraph from a photographed page into a document.
  • Search a screenshot for a word or an order number.
  • Pull the lines from a slide, a poster or a sign.
  • Keep a text copy of a receipt or a letter you only have as a picture.
  • Check a quote against the original image before you share it.

Remember that a picture of a book page or a paid report may be protected by copyright. Use the text only in ways the owner allows.

Photos of documents

A phone photo of a page often has a tilted view, a curved page and uneven light. Lay the page flat, hold the phone parallel to it and use daylight rather than a flash, which leaves bright patches. If the page is long, take one picture per section so that the letters stay large. A scan from a flatbed scanner at 300 dots per inch is usually the best input. When a picture is mostly photograph with only a small caption, crop to the caption first, because the engine then spends its effort on the words.

Tips for better text

  • Use dark text on a plain light background when you can.
  • Zoom in or take the photo from straight above the page.
  • Check numbers and names carefully; one wrong digit is easy to miss.
  • Run the text through a spell checker if it will be published.
  • For text inside a PDF that you can select, copy it directly instead.

This tool is not affiliated with Google or the Tesseract project.

What next

Use this often? Add Image to Tex to your Home screen.

Frequently asked questions

How does image to text OCR work in the browser?

A WebAssembly version of the Tesseract engine reads your picture on your own device and returns the words it finds. Nothing is uploaded.

Which languages does it read?

English only. Other languages need larger data files, so they are not included here.

Can it read handwriting?

It can try, but handwriting is read much less accurately than printed text. Clear typed or printed text gives the best results.

Why is the first picture slow?

The engine and language data, about 8 MB, load the first time. After that the browser can reuse them, so later pictures start faster.

How accurate is the text?

It depends on the picture. The page shows the reader confidence score. Always check the text against the picture, especially numbers and names.

Can I read a PDF with it?

Not directly. Take a screenshot or export a page as a picture first. If the PDF already has selectable text, copy it instead.

What went wrong?

We only receive the reason you pick, never what you typed into the tool.

Thanks for your feedback.

More Image Tools