Turn photos and scans into an editable Word document with OCR in your browser: 30 languages, unsure words marked, nothing uploaded
Add the picture to JPG to Word, choose the language of the text and click Convert to Word. The OCR engine reads the text in your browser and you download a .docx with editable paragraphs. Words it is unsure of are highlighted in yellow so you can check them. Nothing is uploaded.
Optical character recognition (OCR) finds the letters in a picture of text and turns them into real text, which is then saved as editable paragraphs in a Word document.
Related terms
Turn photos or scans of text into an editable Word document with OCR, in your browser.
Add your pictures
Drop one or more JPG, PNG, WebP or HEIC files and put them in order.
Choose the language
Pick the language of the text; English and Hindi load from AnyTool.
Choose what goes in the file
Editable text, the picture with its text, or the picture only.
Convert and check
Click Convert to Word, then check the words highlighted in yellow.
Download
Save the .docx and open it in Word, Google Docs or LibreOffice.
Result: An editable Word (.docx) document
The text becomes normal Word paragraphs you can edit, copy and format. Larger lines, such as titles, become headings.
The Tesseract OCR engine runs in your browser. English and Hindi need nothing from outside AnyTool; for other languages the browser downloads that language's data file once from the jsDelivr CDN. The picture itself never leaves your device.
OCR reads the words and paragraphs, not the design: columns, tables, fonts and colours are not rebuilt. You can put the picture above the text in the document to compare. For tables, use Image to Excel.
Sharp, straight, well-lit photos and 300 dpi scans of printed text read well. Blurry, angled or low-contrast pictures and handwriting read poorly. Words the engine is less than 60% sure of are highlighted in yellow, and the average confidence is shown.
English, Hindi, Spanish, French, German, Portuguese, Arabic, Urdu, Bengali, Tamil, Chinese (Simplified and Traditional), Japanese, Korean, Russian and more. Choose the language of the text before converting.
Add several pictures and put them in order; each starts on a new page of the same Word file, if you like.
Choose 'Picture only (no OCR)' and each picture is placed on its own page of a .docx.
Each picture is decoded in the browser (iPhone HEIC photos included, turned upright from their rotation tag) and scaled so text is large enough to read. Tesseract (tesseract.js 7) finds the paragraphs and lines; each paragraph becomes a Word paragraph, and lines at least 1.45 times taller than normal become headings.
Every word comes with a confidence score. Words under 60% are highlighted in yellow in the Word file, and the page shows the average confidence. Clear printed text usually scores above 85%.
It reads text, not layout: columns, tables, fonts and colours are not rebuilt, and handwriting is rarely read well. TIFF files cannot be opened by browsers and must be converted to PNG or JPG first.
The OCR engine runs in the browser and English and Hindi are served by AnyTool. Other languages fetch only their language data file from the jsDelivr CDN; the picture is never uploaded.
| In the picture | In the Word file |
|---|---|
| Printed text | Editable paragraphs |
| Titles in larger type | Word headings |
| Line breaks | Joined into paragraphs, or kept |
| Tables and columns | Text only (use Image to Excel for tables) |
| Fonts, colours, images | Not rebuilt; the picture can be added above the text |
| Handwriting | Usually not readable |
Accuracy depends on the picture: sharp, straight, evenly lit text reads best.
