How to Extract Text from an Image on Any Device

There is no single right way to extract text from an image. What works depends on what you have in your hand: a phone with the photo already in the camera roll, a work laptop where you cannot install anything, a file someone emailed you. This guide walks through each route, says plainly where it falls short, and tells you which one to reach for first.

If you just want the fastest possible copy-paste, read how to copy text from an image instead — it is the same job, stripped to one path. If you care about the output format (Word, Excel, Markdown), see how to convert an image to text.

The short answer

Open the copy text from an image tool, drop the picture in, and copy the result. It runs the OCR model inside your browser, which means it works on any operating system, needs no install and no account, and never uploads the file. Everything below is context for when you would pick something else.

On an iPhone or iPad

iOS has Live Text built in. Open the photo, press and hold on any text in it, and the usual selection handles appear. Drag them, tap Copy.

This is the fastest option when the image is already in Photos, and the recognition quality is genuinely good. Its limits show up quickly though: it only works on images the Photos or Camera app can open, it will not touch a PDF page, and there is no way to pull a table out as a table. If the text is small or the photo is at an angle, Live Text often selects half a line and refuses the rest.

On Android

Google Lens is the equivalent, reachable from the Photos app or the Lens icon in the camera. Point it at the image, tap Text, then Copy text.

Lens is strong on photographs of signs and documents and it will happily translate as it goes. The catch is that it sends the image to Google’s servers — which is fine for a menu, and worth thinking about for a payslip or a contract. It also tends to reflow multi-column layouts into one stream, so a two-column page comes back interleaved.

On Windows

Two built-in options, both underused:

• Snipping Tool. Take a capture, then use the text actions button in the toolbar to copy the recognised text straight out of the snip. This is the quickest path when the text is on your screen right now rather than in a file.
• PowerToys Text Extractor. If you have Microsoft PowerToys installed, Win + Shift + T turns the cursor into a region selector and copies whatever text is inside the box. It is the closest thing Windows has to a system-wide OCR shortcut.

Both are screen-based: they read what is displayed, not what is in a file on disk. To extract text from an image file, you still have to open it first.

On a Mac

macOS has the same Live Text engine as iOS. Open the image in Preview or Quick Look, select the text with the cursor, copy. It works well on clean scans and screenshots and struggles with the same things iOS struggles with — angles, low contrast, dense small type.

In the browser, on anything

This is the route that does not care what you are running. Drop the file on the copy text from an image page and the recognition happens in the tab:

1. Drop, paste or pick the file. JPG, PNG, WebP, GIF, BMP and AVIF all work. A screenshot pasted with Ctrl + V works too.
2. Extract everything, or one region. Drag a box on the preview to take a single paragraph instead of the whole poster.
3. Copy it out. Plain text, or a rebuilt table you can paste into a spreadsheet.

Because the model runs locally, the file never leaves your device — the reasoning behind that choice is in why OCR should run in your browser. It also keeps working with the network off after the first load.

Which one should you actually use?

If the image is already on your phone and you need one line, use the built-in tool — Live Text or Lens will be faster than anything else. If you are on a computer, if the file is a document rather than a snapshot, if it contains anything you would not paste into a stranger’s website, or if you need a table to survive as a table, use the browser tool.

When extraction goes wrong

Nearly every failure traces back to the input rather than the engine. Small text is the most common cause: lowercase letters need roughly 20 pixels of height before any OCR model is reliable, so get closer or zoom in before capturing rather than upscaling afterwards. Angled shots are the second — photograph pages square-on, not from the side of the desk. Low contrast is the third: faded receipts and grey-on-grey UI text are hard for models trained mostly on black-on-white.

There is a fuller checklist in 7 ways to get better OCR results. And if the text refuses to come out because a website is blocking you rather than because it is an image, that has its own guide.