How to Extract Text From an Image: Every Device, Step by Step

Every mainstream platform has optical character recognition (OCR) built in, and each hides it somewhere different. Here is where it lives on each surface, checked against each vendor’s own support pages, plus the browser route when you cannot install anything.

iPhone and iPad: Live Text

Apple’s support page on copying text from photos is explicit about the hardware: Live Text needs an iPhone XS, iPhone XR or later with iOS 15 or later, or iPadOS 15.1 or later on the iPad models Apple names.

Open Photos, touch and hold a word, then drag the grab points to widen the selection. Tap Copy, or Select All for the whole frame. You can also skip the photo: point the Camera at a page and tap the Live Text button in the viewfinder. Apple states Live Text is not available in every region or language, so test an unusual script.

Android: Google Lens in Photos

Google’s Photos help page for Android gives three steps: open Google Photos, select a photo, then tap More and Google Lens. To copy, tap and select the words you want. The same button sits in the Google app and most camera apps, working on the live viewfinder.

One caveat: Google’s help page documents copying on the phone, not a copy-to-computer step, and threads in Google’s own forums show that button coming and going between releases. If the text has to reach a desktop, plan on moving it yourself.

Windows: Snipping Tool, then PowerToys

Microsoft’s Snipping Tool page, marked as applying to both Windows 11 and Windows 10, is direct: take a snip, then select the Text actions button to activate OCR. You can select specific text, use Copy all text, or use Quick redact to blank out email addresses and phone numbers. Microsoft states that all text recognition is performed locally on your device.

If your build has no such button, PowerToys ships Text Extractor. Microsoft Learn gives the default shortcut as Win+Shift+T: drag a rectangle and the text goes to the clipboard. It recognises only languages whose OCR language pack is installed, and Microsoft warns the text may not be perfect, so proofread it.

Mac: Live Text in Photos and Preview

Apple documents Live Text in both. In Photos, put the pointer over the text, drag to select, then Control-click and choose Copy Text, or press Command-C. Preview is identical, and its page carries the fix for when nothing highlights: choose Tools → Text Selection.

Microsoft Word: the honest answer

Word does not run OCR on a picture you paste into a document. No command turns that image into words.

What Word does is convert PDFs. Microsoft’s page on opening PDFs in Word says Word makes a copy of the PDF, converting it to a Word document and attempting to match the layout, and that this works best with files that are mostly text. It states the failure case plainly: if the PDF contains mostly charts or other graphics, the whole page might show up as an image, and then the text cannot be edited. A scan is exactly that case.

The working Office route is OneNote: right-click the picture, click Copy Text from Picture, paste with Ctrl+V. Microsoft lists it for OneNote for Microsoft 365, 2024, 2021 and 2016, warns that some results take 24 to 48 hours to become available, and says effectiveness depends on image quality. One piece of old advice is dead: Microsoft’s retirement notice says Microsoft Lens left the App Store and Google Play on 9 February 2026 and stopped creating new scans on 9 March 2026.

Google Docs: the conversion happens in Drive

Drive does the OCR and Docs shows the result: upload the file, right-click it, then choose Open with → Google Docs. Google’s conditions are specific and worth reading before you blame the tool:

• PDFs, or photo files in .jpeg, .png and .gif
• The file should be 2 MB or smaller
• Text should be at least 10 pixels high
• Documents must be right side up; rotate before uploading
• Common fonts such as Arial or Times New Roman work best
• Use sharp images with even lighting and clear contrasts

Google adds that bold, italics, font size and line breaks are likely to be retained, while lists, tables, columns and footnotes are not. Expect words, not layout.

Excel and Illustrator

Excel’s Data from Picture is real but narrow. It is a Microsoft 365 feature; on Windows, Microsoft requires Windows 11 or Windows 10 version 1903 or later with the Edge WebView2 Runtime. It reads a table into cells, not paragraphs into text. The character sets Microsoft lists are 21 Latin-script European languages plus Turkish, and it wants a head-on image of the data alone.

Illustrator has no OCR. Image Trace converts pixels into vector paths that can look like letters but are shapes, not editable type. Adobe’s own Image Trace guidance says so and sends you to Acrobat, where All tools → Scan and OCR → Recognize Text adds a searchable text layer.

In a browser, with nothing installed

On a borrowed machine, a locked-down work laptop, or a phone whose built-in OCR does not cover your language, use image to text. It runs Tesseract compiled to WebAssembly inside the page, offers 21 selectable languages, and takes JPG, PNG, WebP and BMP. The result lands in an editable box, so you can fix the inevitable stray character before copying it or saving a .txt file. The image never leaves your device; the only download is the language file. HEIC photos from an iPhone need converting to JPG first.

Why OCR gets it wrong

Every engine above fails on the same material, so if one chokes, the next usually will too:

• Blur — a handheld shot at the wrong shutter speed. Retake it rather than retry it.
• Skew — a page photographed at an angle. Google states outright that documents must be right side up.
• Low contrast — grey text on a grey scan, or a shadow across half the page.
• Tiny type — Google’s floor of 10 pixels per character is a fair rule everywhere. Enlarging a small file adds no detail that was never captured.
• Handwriting — not reliable in any of these tools. Neat block capitals sometimes survive; cursive generally does not.

Screenshots are the easiest wins by a wide margin: perfect focus, no skew, real pixel edges. If you are unsure how large an image truly is, check its EXIF data for the pixel dimensions before blaming the OCR. And a QR code is not text at all — no OCR will read one; point a camera at it, or generate one of your own.

Check the PDF before you OCR it

A PDF that already contains a text layer needs no OCR. Drag your cursor across a line; if words highlight, copy them straight out — Apple documents this for Preview as Tools → Text Selection, drag, then Edit → Copy. Recognising that file again only replaces good text with a worse guess. Only a scanned PDF, a photograph of pages, needs OCR, and browser tools take images rather than PDFs, so convert the pages to JPG first.