Map Extract
Free tool

Extract text from an image

Photograph a flyer, a price board or a trade stand banner and turn the writing into text you can copy. Recognition runs on your own device and gives a confidence score.

Runs in your browser. On first use it downloads an open-source processing engine from a public code CDN; your files stay on your device.

Reading text from a picture

  1. Choose the image

    Press Choose a photo, scan or screenshot.

  2. Set the language

    Pick the language of the text: English, French, German, Spanish, Italian, Portuguese, Dutch, Polish, Welsh or Irish.

  3. Extract

    Press Extract text. The first time you use a language, the engine and that language's data download, and a progress line shows each stage.

  4. Check and keep

    You get the number of words found and a confidence percentage. Correct anything in the text box, then Copy it or Download .txt.

Example: contact details from an exhibition stand

At a regional trade show, a wholesale buyer photographs the banners of a dozen small food producers instead of collecting leaflets. Back at the office, each photo goes through OCR in English.

For a banner showing a name such as Hillside Preserves, an address like sales@hillside-preserves.example and a number like 0161 496 0000, the buyer checks the recognised text against the photo, corrects any misread characters, and pastes it into the email and phone extractors to pull out clean contact fields for a follow-up list.

Input
stand-banner.jpg, language English

For a better result

When confidence comes back under 70%, the tool suggests a better photo. These help most:

  • Shoot straight on, not at an angle, and fill the frame with the text.
  • Use even light and avoid glare on glossy signs or laminated menus.
  • Crop away everything that is not text before extracting.
  • Choose the right language; accents and letters such as Welsh ŵ and ŷ depend on it.
  • Check numbers and email addresses character by character: 0 and O, 1 and l are easily confused.

What OCR cannot do

Handwriting, decorative script fonts and text curved around a logo are hard for any OCR engine. Columns and tables are read line by line, so layout is not preserved. Treat the result as a draft to proofread, not as an exact copy.

Questions people ask

How do I copy text from a picture?

Load the picture here, choose the language of the text and press Extract text. When recognition finishes, the text appears in an editable box. Fix any mistakes, then use Copy or Download .txt. Screenshots of clear on-screen text usually come out almost perfectly.

Is my image uploaded for OCR?

No. The open-source recognition engine and the language data are downloaded from a public code library the first time you use a language, then cached by your browser. The image itself is read on your device and never sent anywhere.

What does the confidence score mean?

It is the engine's own average certainty about the words it recognised, from 0 to 100%. High scores usually mean clean text; below 70% expect errors, typically from blur, skew or low contrast. It does not guarantee every character is right, so always proofread.

Where your data goes

Text recognition runs in your browser. On first use, the open-source tesseract.js engine and the chosen language data (a few megabytes) are downloaded from the cdn.jsdelivr.net public code library; your image is not sent with them or anywhere else.

Checked against the tool on

Top