How do I convert a JPG to a Word document?

Choose, drop or paste the JPG, turn it upright if it is sideways, and crop it to the text if the picture holds more than you need. Press Read text and the words appear in the review pane, where you can correct them. Download under Word document then saves a .docx with one paragraph for each line of the pane, on A4 pages with one-inch margins.

Is my image uploaded anywhere?

No. The image is decoded and read in your browser tab by Tesseract, whose code and English model this site serves as fixed files. Nothing about the image or its text is sent to this site or to anyone else. The only things saved, on your device, are your line break and word box settings.

Does it keep the layout, tables and fonts of the original?

No. It reads text, not design: a table comes out as lines of words, and fonts, sizes, colours and pictures are not carried over. Choose As printed to keep every printed line break, or Paragraphs to join each paragraph's lines into one, which is usually easier to edit in Word.

Can it read handwriting or languages other than English?

No. The model is Tesseract's fast English model for printed text. Handwriting and other languages come out wrong or not at all, so check anything like that word by word, or use a tool built for it.

Why are some words outlined in red?

Every word Tesseract finds gets a box on the image. Words it scored below 60 out of 100 are outlined in red, and the Word boxes line names the lowest-scoring word, so you know where to look first. A high score is not a promise that the word is right, so read the rest too.

Which files work, and is there a size limit?

JPG and PNG, recognised by their contents rather than their names. A HEIC photo, WebP, GIF, BMP, TIFF or PDF is named and turned away with what to do instead; for a PDF, PDF to Image turns its pages into JPG or PNG files first. There is no file size limit, but anything over 12 megapixels is scaled down to fit 12 before reading, and so is an image with a side longer than 16,384 pixels, so crop a large photo to the text to keep the words sharp. The first read in a tab also loads the OCR engine, 8.1 MB, from this site.

Image to Word

Image to Word reads the printed English in a JPG or PNG and puts it in an editable review pane, with a box drawn around every word it found on the image and the words it is least sure of outlined in red, so you can check and correct the text before you download it as a Word document or a plain text file. You can turn the image in quarter turns and crop it to the text first. The OCR engine and its English model are files this site serves, and they run in your browser: the model is checked against a pinned checksum before any text is read, the image never leaves your device, and there is no account. Anything over 12 megapixels is scaled down to fit 12 before it is read.

Drop a JPG or PNG here

, paste one, or

Never uploadedNo accountEngine served by this siteEdit before export

Image

none yet

Turn

Crop

Line breaks

Choose, drop or paste a JPG or PNG, turn it upright and crop it to the text if you need to, then press Read text. Correct any word in the review pane before you export: reading again, or choosing another image, replaces that text.

Word document.docx, one paragraph per line
nothing to export yet
Plain text.txt, UTF-8
nothing to export yet
Clipboardthe review text, as edited
nothing to copy yet
Word boxesoutlined on the image
after the first read

Read in this tab, with files this site serves. The engine is tesseract.js 7.0.0 with tesseract.js-core 7.0.0, LSTM build without SIMD, and the model is tessdata_fast English (eng.traineddata), SHA-256 beginning 7d4322bd2a77, retrieved 2026-09-21. The model is checked against that pinned checksum before any text is read, and a mismatch stops the tool instead of guessing. The image is never uploaded; only your line break and word box settings are remembered on this device. Credit: Tesseract.js and tessdata_fast, both under the Apache License 2.0 (engine licence, model licence, bundled notices).

Common questions

How do I convert a JPG to a Word document?: Choose, drop or paste the JPG, turn it upright if it is sideways, and crop it to the text if the picture holds more than you need. Press Read text and the words appear in the review pane, where you can correct them. Download under Word document then saves a .docx with one paragraph for each line of the pane, on A4 pages with one-inch margins.
Is my image uploaded anywhere?: No. The image is decoded and read in your browser tab by Tesseract, whose code and English model this site serves as fixed files. Nothing about the image or its text is sent to this site or to anyone else. The only things saved, on your device, are your line break and word box settings.
Does it keep the layout, tables and fonts of the original?: No. It reads text, not design: a table comes out as lines of words, and fonts, sizes, colours and pictures are not carried over. Choose As printed to keep every printed line break, or Paragraphs to join each paragraph's lines into one, which is usually easier to edit in Word.
Can it read handwriting or languages other than English?: No. The model is Tesseract's fast English model for printed text. Handwriting and other languages come out wrong or not at all, so check anything like that word by word, or use a tool built for it.
Why are some words outlined in red?: Every word Tesseract finds gets a box on the image. Words it scored below 60 out of 100 are outlined in red, and the Word boxes line names the lowest-scoring word, so you know where to look first. A high score is not a promise that the word is right, so read the rest too.
Which files work, and is there a size limit?: JPG and PNG, recognised by their contents rather than their names. A HEIC photo, WebP, GIF, BMP, TIFF or PDF is named and turned away with what to do instead; for a PDF, PDF to Image turns its pages into JPG or PNG files first. There is no file size limit, but anything over 12 megapixels is scaled down to fit 12 before reading, and so is an image with a side longer than 16,384 pixels, so crop a large photo to the text to keep the words sharp. The first read in a tab also loads the OCR engine, 8.1 MB, from this site.

Printed English is read by a pinned Tesseract model in your browser, and OCR can still misread characters, most often in photos and small print. It does not read handwriting, rebuild tables or keep the page layout, so check the text against the image before you rely on it. The Word and text files hold exactly what is in the review pane.