Image to Text Converter

Optical character recognition, or OCR, turns the shapes of printed letters in a picture into editable text. Here it runs on your own device, so a private letter or a photographed receipt is not sent anywhere.

1. Choose the Language in the image: English, Urdu, Arabic, Hindi, Spanish, French, German, Portuguese, Turkish or Russian. Pick the one language that the text is written in.
2. Leave Keep line breaks as in the image ticked to preserve every line as it appears, or untick it to join the lines of a paragraph into flowing text.
3. Drop one image into the drop zone, click to browse, or paste a screenshot. JPG, PNG, WebP and other formats your browser can open are accepted.
4. Press Extract text. A progress note shows the stages, from loading the engine to reading the text.
5. Check the text in the result box against the thumbnail of your picture, then use Copy or Download to save a .txt file.

The first run downloads the OCR engine (about 4 MB) and the data for your language (about 1 to 3 MB) from this site, which takes longer on a slow connection. The browser caches both, so later runs start faster.

The engine is Tesseract, an open-source OCR system that began at HP and was later developed at Google. Before reading, the page redraws your picture: a small image is enlarged up to 3 times, a very large one is reduced to 3,600 pixels on the longest side, transparency is flattened onto white, and the rotation that phones record is applied. Two examples show the effect:

Enlarging helps the engine but cannot add detail that was never captured, so size alone does not rescue a blurred picture. What matters is the height of the letters in the original: aim for text at least about 20 to 30 pixels tall. As a worked case, 10 pt text on a page scanned at 300 dpi is about 42 pixels tall, which reads well, while the same page scanned at 100 dpi gives 14 pixel letters and many mistakes. If your source is a PDF, see the next section for the right resolution.

• Helps: sharp, well-lit, straight-on pictures, dark text on a plain light background, and one language per run.
• Hurts: handwriting, decorative fonts, tiny or blurred text, low contrast, curved pages and text laid over a photograph.
• Fix first: straighten a sideways picture with the rotate and flip tool, and cut away borders and photos with the image cropper.

The result line reports the words, the characters, an "engine confidence" percentage and the time taken. Confidence is the engine's own estimate, not a guarantee, so a high figure can still hide a wrong digit.

OCR is most useful for short, clean passages. Copy a paragraph from a screenshot of a page that blocks selecting text, with Keep line breaks off so the lines join into one paragraph you can paste into an email. Digitise a printed letter or a recipe card, pull a phone number or reference from a photographed sign, or lift quotations from a scanned book page. Once you have the text, the word counter can measure it and the plain text converter will strip stray formatting.

Know the limits before you rely on it:

• Proofread everything. Check numbers, names, dates and amounts against the picture, because 1, l and I, or 0 and O, are easy to confuse.
• Layout is lost. Fonts, tables, columns and formatting are not kept, and in multi-column pages or tables the reading order may be mixed.
• Scripts vary. English and the other Latin-script languages read most reliably. Arabic and Hindi work on clean printed text. Urdu in the flowing Nastaliq style often reads poorly, because the model is aimed at simpler print.
• No handwriting. Neat handwriting may give a few words, but expect poor results.
• One image per run, and no PDFs. To read a PDF, first turn its pages into pictures with the PDF to JPG converter, choosing PNG at Print (216 dpi) so that 12 pt text is about 36 pixels tall, then extract each page in turn.

If nothing is found, the tool says so. Try a sharper picture, darker text on a plain background, and check that the language setting matches.