imgglide
English

Extract text from an image

Point it at a screenshot, a snapshot of a sign or a scanned page and it types out the words for you. The reading happens in your browser, so the picture stays where it is.

Add one image or PDF that contains text.

or drop a file here

How to use it

  1. Add one picture or PDF with ‘Choose a file’, or drop it on the page. Photos, screenshots and scans all work, and a PDF is read page by page.
  2. Under ‘Document language’, select every language the text is written in. Each one is named in its own language, such as Deutsch or Español. English is selected to begin with, and your choice is remembered for next time.
  3. For a PDF, decide between ‘All pages’ and ‘Page range’; with a range, type the pages you want, such as 1-3, 5.
  4. Click ‘Extract text’. The first run downloads the reading engine and the data for your languages; a progress bar follows, with a line like Reading page 2 of 5 for PDFs.
  5. Fix wrong letters right in the result box, then click ‘Copy’ or ‘Save as TXT’. ‘Start over’ clears the page for the next file.

What it does

  • Pictures up to 3000 pixels

    A picture is read at up to 3000 pixels on its longer side. Anything bigger is scaled down first, because reading gets slower without getting more accurate. Transparent areas are filled with white before the reading starts.

  • Scanned PDFs, page by page

    Each page is drawn at twice its normal size, which is 144 dpi, and read in turn. Pick all pages or a range; when more than one page is read, the text of each page starts under its own page heading.

  • Seven languages, mixed if you like

    English, German, French, Spanish, Portuguese, Japanese and Korean. Select two or three for a bilingual menu or sign, but know that each extra language makes the reading slower.

  • A result you can edit

    The text lands in a box you can type in, so you can fix misread letters before you take it anywhere. ‘Copy’ puts it on the clipboard, and ‘Save as TXT’ downloads a plain text file named after the original.

  • A count after every run

    When the reading ends, a line reports how many characters were found, spaces and line breaks not included. If there are none, it suggests a sharper image or a different document language.

  • Engine and languages load once

    The reading engine is roughly 4 MB, and each language adds between 0.7 and 3 MB. They come from this site on the first run and your browser keeps them. The picture itself never leaves your device.

Good to know

OCR, short for optical character recognition, turns the letters in a picture into real text. It earns its keep whenever words are stuck inside an image: an error message in a screenshot, a printed receipt, a page from a book, a sign you photographed, a scanned letter. Instead of retyping, you copy.

The picture matters more than any setting. Get close so the text fills the frame, lay the page flat, keep shadows off it and avoid blur; dark letters on a light background read best. Handwriting, tiny print, decorative fonts and tables with merged cells are the weak spots, and the result is plain text, so bold type, columns and table layout are not kept.

Every PDF page is drawn as a picture and then read, so for a PDF whose text you can already select, copying it straight from a PDF viewer is more exact. Recognition can still swap similar shapes, such as 0 and O or l and 1, so check names, numbers and amounts against the original before you rely on them.

Questions and answers

How do I copy text from a picture?

Add the picture, select the language of the text, click ‘Extract text’, and use ‘Copy’ on the result box.

Which languages can it read?

English, German, French, Spanish, Portuguese, Japanese and Korean. Select every language the document uses; each extra one slows the reading down.

Can it read a long PDF, or only some pages?

Both. Choose ‘All pages’, or ‘Page range’ and type something like 1-3, 5. Fewer pages finish much sooner.

Why are some letters wrong?

The recognition is automatic. Blurry, tilted or low-contrast pictures cause mistakes. Correct the text in the box, or shoot the picture again from nearer, in better light.

Can it read handwriting?

Not reliably. The engine is built for printed letters, so neat block writing may come out partly right and cursive usually will not.

Is my picture uploaded?

No. The first run fetches the reading engine and the language data from this site; after that the recognition itself happens in your browser.

Related tools