Images

Extract text from an image

Choose a photo of a page, a screenshot or a scanned PDF, and the text appears right here, ready to fix, copy or download as .docx. The picture never leaves your device.

  1. Choose a picturePick a photo of a page, a screenshot or a scanned PDF. Nothing is uploaded.
  2. Recognize the textCheck the document language and tap Recognize text; the text appears on the page.
  3. Copy or downloadFix anything misread, then tap Copy text or Download .docx.

Save and continue on the desktop

Download or copy your result here first. Open the free desktop to save to Files or use more features.

  • Edit Word files, spreadsheets and slides
  • Save to Files when you continue on the desktop
  • Every tool in one place, no new tabs
Open the free desktop

The full Documents editor is already on the desktop: chapters, tables, checklists. The recognized text opens there in one click.

What Image to Text can do

For the letter that arrived as a picture

An official letter sent as a JPEG, a screenshot of a bank notice, a photo of a contract page: the words are there, but none can be copied. Choose the picture, and the text comes up here as an editable document. Fix a misread letter, then copy it or download a Word file.

The reading model downloads once

The first time, recognition downloads its reading model, about 5 MB. After that it works offline, even with no signal.

Seven languages

It reads English, Russian, Spanish, German, French, Turkish and Indonesian. The page’s own language is picked for you, and “Document language” switches it before reading. English is always read alongside, so e-mail addresses and company names in Latin letters survive in any of them.

What it reads well, and what it does not

Printed text reads best: a clean scan or a screenshot. In our tests up to about 6 characters in 1,000 came out wrong on scanned letters, and 1 to 2 in 100 on phone photos. Bold and italic come out as plain text, pictures are left out, and handwriting is not read; a table with clear columns comes through as a table, ready for Excel. Check names, numbers and dates.

Related apps

All AI tools

Questions

Is there a limit on how many pictures?

No. Recognition runs on your own device, so there is nothing to count. Read as many as you need.

Is my picture uploaded anywhere?

No. It is read in your browser, on this device, and never sent to a server. Only the reading model is downloaded, once.

Which languages does it read?

English, Russian, Spanish, German, French, Turkish and Indonesian. Pick the document’s language before reading; English is read alongside it.

Can it read handwriting?

No, printed text only. Handwritten notes and signatures are skipped or come out as noise.

Can I get a Word file instead of plain text?

Yes. Under the recognized text there is “Download .docx” next to “Copy text”. Paragraphs, headings and lists come over; you can edit the text on the page first.

Can I get a table into Excel?

Yes, when its columns line up: an invoice, a price list, a statement. Under the recognized text, “Download the table (.xlsx)” writes the cells into a spreadsheet, amounts as numbers you can add up. Merged cells come through as text; check the figures against the picture.

What about a scanned PDF?

Choose it the same way. Its first 20 pages are read; split a longer scan and read it in parts. For a PDF that already has text in it, the PDF to Word page keeps the bold, italic and pictures too.