How to Copy Text from an Image or Screenshot (Free OCR)

The fastest way: paste a screenshot

Screenshots are the easiest images to read because the text is sharp, straight and evenly lit.

1. Take a screenshot. On Windows, press Win + Shift + S and drag over the text. On a Mac, press Cmd + Shift + Ctrl + 4 to copy a selected area to the clipboard. On a phone, take a normal screenshot.
2. Open Image to Text and press Ctrl+V (or Cmd+V). On a phone, tap Choose images and pick the screenshot.
3. The tool reads it straight away. The first run downloads the reading engine (about 1.4 MB) and your language data, so it takes a little longer.
4. Check the text, fix anything that's wrong, and press Copy text.

You can add several images at once. Each gets its own result, and Copy all copies the text from every image in one go. All as .txt saves everything in a single text file.

Choosing languages: Hindi, English and more

Under Text language, tap every language that appears in the image. The tool supports English, Hindi, Bengali, Marathi, Tamil, Telugu, Gujarati, Kannada, Malayalam, Punjabi, Urdu, Arabic, Spanish, French, German, Portuguese, Russian, Chinese (Simplified) and Japanese.

A few rules make a big difference:

• Pick the right script. If Hindi is not selected, Devanagari text comes out as random Latin letters, or nothing at all.
• Mix up to three languages. For a Hindi form with English headings, select both. The first language you pick is treated as the main one.
• Don't add extras "just in case". Each extra language makes reading slower and can confuse similar-looking letters. That's why the limit is three.

Language files download once and stay in your browser. Hindi plus English is about 4.3 MB of language data the first time. The tool tells you how much it will download before it starts.

Tip: Changed the language after reading? A note appears saying settings have changed. Tap Read it again to re-run that image with the new language.

The accuracy settings and when to change them

Under Accuracy, three switches are on by default. Leave them on unless you have a reason not to.

Enlarging helps more than you'd expect. The engine reads best when letters are fairly large. A screenshot 800 px on its longest side is scaled up three times before reading. One 1,200 px long is doubled. Images 2,000 px or larger on their longest side are left as they are.

Getting good results from photos of documents

A phone photo of a printed page can read almost as well as a screenshot, but only if you take it carefully.

• Light: Use bright, even daylight. Avoid the flash, which leaves a shiny hotspot, and avoid your own shadow falling across the page.
• Angle: Hold the phone directly above the page, parallel to it. A slanted photo makes lines at the top smaller than lines at the bottom.
• Flatness: Press books open and smooth out folded letters. Text that curves into the spine is the hardest to read.
• Distance: Fill the frame with the text you need. A whole page shot from far away gives each letter only a few pixels.
• Focus: Tap the screen on the text before you shoot and keep still. Blurry letters can't be fixed afterwards.

On a phone, the Take a photo button opens the camera directly.

If you only need one paragraph from a busy page, crop first. Open the photo in the Image Resizer, use the Crop tab to keep just that part, and download it. Less clutter means fewer stray characters.

Checking, editing and copying the text

After reading, the tool shows a confidence badge: High (85% or more), Medium (60–84%) or Low (below 60%). It's an average for the whole image, so even a High result can have a wrong word or two.

To find errors quickly:

1. Use Highlight on the image and choose Lines or Words. Words the engine wasn't sure about (below 60% confidence) are flagged, so you know where to look.
2. Compare those spots with the image and correct them in the text box. The box is fully editable.
3. If you make a mess of it, the undo button restores the original result.
4. Press Copy text, or .txt to save it as a file.

The word and character count updates below the box as you edit. For longer checks, like reading time or keyword counts, paste the text into our Word Counter.

Common OCR slip-ups to look for:

• 0 and O, 1 and l and I, rn and m.
• Hindi matras (vowel signs) dropped or attached to the wrong letter, especially in small or bold fonts.
• Table columns merged into one line. OCR reads left to right, so tables often need tidying by hand.
• Phone numbers, amounts and account numbers. Always check digits against the original.

Privacy: what happens to your image

The reading happens in your browser using a WebAssembly build of the Tesseract engine. Your image is processed on your device and isn't uploaded to us or anyone else.

Two things do come from the internet: the reading engine itself and the language files. They download once from a public CDN and your browser keeps them, so later runs work faster. That makes the tool reasonable for bank statements, ID cards and medical letters. As with any device, clear your browser data afterwards if you're on a shared computer.

Limits: handwriting, PDFs and tricky images

Handwriting. The engine is built for printed and on-screen text. Neat capitals sometimes work; cursive, notes and signatures almost never do. If you get "nothing readable", that's usually why.

PDFs. The tool accepts images only (JPG, PNG, WebP, GIF, BMP, AVIF and HEIC). If your PDF has selectable text, copy it straight from your PDF reader instead. For a scanned PDF, take a screenshot of each page and paste it. For many pages, use PDF to JPG to turn every page into an image, then add them all at once.

Stylised text. Logos, text over photos, very thin fonts and heavy shadows reduce accuracy. Crop tightly and keep Boost contrast on.

Very low resolution. A thumbnail forwarded several times on WhatsApp may simply not have enough detail. Ask for the original image if you can.

If a result is poor, work through this order. Check the language, crop to the text, and retake the photo in better light. Then try Read it again. Most bad results come from the wrong language or a blurry photo, not the engine.

Everyday uses

Once you're used to it, OCR saves typing in lots of small ways:

• Phone numbers and addresses from a visiting card or a shop sign.
• Quotes and paragraphs from a screenshot someone shared, so you can search or translate them.
• Error messages from a screenshot, so you can paste the exact text into a search engine.
• Notes from slides photographed in a lecture or meeting.
• Old printed letters and certificates you want to retype into a document.

For anything official, such as a PAN, Aadhaar or bank account number, treat the OCR text as a draft. Check every character against the original before you use it.