What image to text conversion is
Image to text conversion is known as OCR, short for optical character recognition. A computer looks at the shapes in a picture, works out which ones are letters, and turns them into text you can select, edit, search and copy.
It is useful any time text is trapped in a picture: a screenshot of an error message, a photo of a sign, a scanned letter, a slide from a presentation, a quote in an image you were sent. Instead of typing it out, you let the computer read it.
How to use the converter
- Choose the Language in the image. English is selected by default. Spanish, French, German, Italian, Portuguese, Dutch and Hindi are also available.
- Give the tool an image in one of three ways:
- Drop an image file onto the box that says Drop an image here, or choose a file.
- Press Choose image and pick one or more files.
- Copy a screenshot, then press Ctrl+V (Cmd+V on a Mac) anywhere on the page, or press Paste image.
- Watch the progress bar. The first scan takes longer because the reading engine and language data have to download. After that, each image takes a few seconds.
- Read the result in Extracted text. You can edit it right there.
- Press Copy, or Download to save it as image-text.txt.
If you choose several images at once, they are read in order and the text from each one is added below the last, separated by a blank line. The counts under the text show the number of Words and Characters, and the average Confidence for the batch.
What works well and what does not
| Image | What to expect |
|---|---|
| Screenshot of a web page, app or document | Very good. Clean, sharp, high contrast text is the easiest kind to read. |
| Photo of a printed page, taken straight on in good light | Good. Check numbers and punctuation. |
| Scanned letter or form | Good for typed text. Boxes and lines on forms can add stray characters. |
| Photo taken at an angle, in shadow or out of focus | Mixed. Crop and retake the photo if you can. |
| Text over a busy photo or a pattern | Poor. The background confuses the engine. |
| Handwriting | Poor, except for very neat block letters. |
| Very small text, such as a full page in a small screenshot | Poor. Zoom in and take a closer screenshot. |
How it works
The page uses Tesseract, a well known open source OCR engine first developed at Hewlett-Packard and later released to the public. It runs here through Tesseract.js, a version of the engine built for web browsers.
When you start the first scan, your browser downloads three things from cdn.jsdelivr.net: a small helper script, the engine itself, and a data file for the language you picked. These are program and language files only. Your browser keeps them in its cache, so later scans start much faster. Choosing a different language downloads that language’s data file.
The engine then finds lines of text in your image, splits them into words and letters, and compares the shapes with what it learned from its training data. All of this happens on your computer or phone. The image is read from your device’s memory and is never uploaded, which makes the tool safe for screenshots that contain private information.
Tips for better results
- Crop to the text. Remove borders, photos and icons around the text before you drop the image.
- Bigger is better. Large, sharp letters read much better than tiny ones. Zoom in on the page before taking a screenshot.
- Pick the right language. Accented letters such as é, ñ and ü are read far more accurately when the matching language is selected.
- Straighten photos. Hold the camera directly above the page, fill the frame with the text and avoid shadows.
- Clean up afterwards. Text from a photo often has line breaks in the middle of sentences. Paste it into the line break remover to join the lines, and use the plain text converter to fix odd quotes and spaces.
- Proofread. OCR mistakes look like real words, for example “rn” read as “m” or “0” read as “O”. Read numbers, names and email addresses twice.
Phone photos saved in the HEIC format cannot be read by most browsers. Save or share the photo as JPG first.
Other ways to get text from an image
Google Docs. Upload the image to Google Drive, right click it and choose Open with, then Google Docs. The text appears below the image in a new document. This uploads the image to Google.
Microsoft OneNote and Word. In OneNote for Windows, paste the image, right click it and choose Copy Text from Picture. Word can open a PDF and convert it to an editable document, which works best for clean scans.
Your phone. Recent versions of iOS and Android let you select text directly in photos and screenshots. Press and hold on the text in the Photos or Gallery app.
Code. Tesseract can be installed on Windows, macOS and Linux and run from the command line with tesseract image.png output, which saves the text to output.txt. Python programs often use it through the pytesseract package.