Skip to content
ImageDoctor
Guide6 min read

How to copy text from an image on any device

Your phone or computer can probably already do it: copying text out of a picture is called OCR, and it is built into iPhone, Mac, Windows and Android. When it can't, or gets it wrong, the fix is usually the image rather than the software.

Processed locally in your browser

To copy text from an image, use text recognition (OCR, optical character recognition). It is built into iPhone and Mac (Live Text), Windows 11 (Snipping Tool's Text actions) and Android (Google Lens), and browser tools such as ImageDoctor's OCR Text Extractor do it without installing anything. Whichever you use, read the result through before relying on it: OCR is very good on clear printed text and unreliable on small, blurry or handwritten text.

Copy text from an image on your phone or computer

DeviceWhereHow
iPhoneCamera (Live Text)Point the camera at the text. When a yellow frame appears around it, tap the Live Text button, then touch and hold the text and drag the grab points to select it
MacPreview (Live Text)Open the image, drag across the text, Control-click the selection and choose Copy Text. If you can't select it, choose Tools › Text Selection
Windows 11Snipping ToolCapture the area, select Text actions, then copy a selection or choose Copy all text
AndroidGoogle Photos or the Google app (Lens)Open the image with Lens, tap and select the words you want, then copy them
Any browserAn online OCR toolAdd the image, wait for the text, copy it — see below

Sources, all checked on 4 November 2026: Apple's Live Text on iPhone and Live Text in Preview pages, Microsoft's Snipping Tool page and Google's Photos and Lens help. Menus move between versions, so if a step doesn't match your device, those pages are the place to check.

Two details are worth knowing. Apple says Live Text isn't available in every region or language. Microsoft says Snipping Tool's text recognition runs locally on your device. Google Lens is a Google service, and its help page explains how to find and delete Lens activity saved to your Google account.

Copy text from an image in your browser

When the built-in options aren't available — an older computer, a language Live Text doesn't cover, or a work machine where you can't install anything — the OCR Text Extractor does it in a browser tab. The recognition engine (Tesseract, compiled to WebAssembly) runs on your device, so the image isn't uploaded to ImageDoctor. The one download is the language model, fetched from a public CDN the first time you use each language and cached by your browser afterwards.

  1. Add the imageDrop in one screenshot, scan or photo. Reading starts straight away in English.
  2. Set the language and layout if neededPick up to three of the 66 languages actually on the page. For layout, Automatic suits a normal page; Single block suits a receipt, label or sign; Scattered text suits words spread over a picture; Single line and Single word suit a code or a title.
  3. Turn the page if it is sidewaysRotate by 90, 180 or 270 degrees. A page that is only slightly crooked is straightened automatically.
  4. Check, edit and copyThe tool reports its confidence and how many words it was unsure about. Correct anything misread in the text box, then copy it or download a .txt file, keeping the line breaks or joining them into paragraphs.

Before reading, the tool also cleans the picture up when Enhance is on, which is the default: it turns it grey, stretches the contrast, inverts light text on a dark background, and enlarges small images (up to four times) towards the resolution the engine was trained on. That is what turns a dim phone photo of a page into something readable.

Getting text out of a PDF

First try selecting the text in your PDF viewer. If it selects, the PDF already contains real text and no OCR is needed. If it won't select, the PDF is a scan — pictures of pages — and you need OCR:

  1. Open the PDF in PDF to Image and render the pages you need as PNG at 300 DPI. PNG avoids JPEG's smudging around letters.
  2. Add each page image to the OCR Text Extractor and copy the text out.

Why OCR gets text wrong, and how to fix it

Almost every misread comes from the image, not the software. The Tesseract project's own advice is that the engine works best on images of at least 300 DPI, that line detection gets much worse when a page is skewed, that uneven background brightness spoils the black-and-white conversion it does internally, and that big empty borders around a single word cause problems (Tesseract documentation, checked 4 November 2026). In practice:

ProblemWhat helps
Text is small in the pictureGet closer, zoom the screen before taking the screenshot, or scan at 300 DPI. As a rule, letters should be at least about 20 pixels tall
Photo is blurry or in shadowRetake it in even light, holding the phone parallel to the page. Sharpening afterwards rarely rescues it
Page is sideways or tiltedRotate it in the tool, or fix the file first with Rotate & Flip. Small tilts are straightened automatically
Lots of other content around the textCrop to the text, leaving a small margin, with the Image Cropper
Mixed or wrong languageChoose the languages actually on the page. An English model reading German drops the accents
HandwritingNot supported. The models are trained on printed text

The resolution point is the one people miss most. A screenshot of a document shown at 50% zoom has letters only a few pixels tall, and no engine can read detail that isn't there. Our explainer on pixels and DPI covers why the pixel count, not the DPI number in the file, is what decides this.

Check the result before you use it

OCR mistakes look plausible, which is what makes them dangerous. The usual suspects are characters that look alike: 0 and O, 1, l and I, rn and m, 5 and S, and commas and full stops in numbers. Check anything where one wrong character matters — account numbers, prices, email addresses, codes and dates — against the original image.

Common questions

Yes. A screenshot is usually the easiest case, because the text is sharp and straight. Use Snipping Tool's Text actions on Windows, Live Text on iPhone or Mac, Google Lens on Android, or add the screenshot to a browser OCR tool.

Some services try, with mixed results. ImageDoctor's OCR Text Extractor doesn't: its models are trained on printed text, so handwriting usually comes out as nonsense.

No. Recognition runs in your browser. The only download is the language model, the first time you use each language; after that it is cached and the tool can work without a connection.

Up to three of the 66 it offers. Each extra language makes it slower and makes look-alike characters more likely to be picked from the wrong alphabet, so choose only the ones on the page.

Tools in this article

All blog