What is the best tool for getting text out of a screenshot?

15 September 2026

There are more screenshot-to-text tools than there are meaningful differences between them. The questions that actually separate them are whether the recognition runs on your machine, and whether the text goes somewhere useful afterwards.

For a one-off, the best tool is the one already on your computer: Live Text in macOS Preview and Quick Look, Text Actions in the Windows Snipping Tool, or Google Lens on Android and in Chrome. All three are free, instant, run locally, and need nothing installed.

For doing it many times a day, a dedicated tool is worth it: TextSniper on macOS and Text Grab or PowerToys Text Extractor on Windows are all fast, keyboard-driven, and local. They put the text on your clipboard and get out of the way.

For capturing text you intend to keep and read again rather than paste once, Second Read is built for that case specifically. It recognises on your own machine, saves the passage with the page it came from, and shows it back to you later on your new tab. If you only ever paste and move on, the clipboard tools above are simpler and you should use one of those.

Is there a free screenshot to text tool?

Yes, and you almost certainly already have one. Live Text on macOS and iOS, Text Actions in the Windows Snipping Tool, Google Lens on Android, and PowerToys Text Extractor on Windows are all free and all included or officially supported.

Among installable free options, Second Read is free with no paid tier, PowerToys is open source and made by Microsoft, and Tesseract is the open-source engine most of the free tools are built on.

Be wary of the free "image to text" websites that dominate the search results. They are free because the image goes to their server, and their business is the traffic rather than the tool.

Do screenshot text tools work offline?

The good ones do. All of the operating system tools listed above run the recognition on your own device with no network involved, as do TextSniper, PowerToys, and Second Read, which runs Tesseract compiled to WebAssembly inside your browser.

Anything that works by uploading, which includes most free websites and some mobile apps, needs a connection by definition and has a copy of your image by consequence.

The test takes fifteen seconds: turn off your wifi and try it. If the text still comes out, the recognition is local and your image never left.

How do you capture text you cannot select?

Text you cannot select has to be read from the pixels, which means taking a picture of the region and running character recognition over it. There is no way to talk the page into giving up text it is not rendering as text.

This covers a surprising amount of what people actually want to quote: Kindle Cloud Reader, embedded PDFs, Google Slides in presentation mode, book and document scans, video frames, screenshots somebody sent you, and any site that has disabled selection.

The capture is the same in every case. Drag a box around the region, let the recogniser read it, and check the result before you keep it. Zooming the page in first materially improves accuracy, because recognition depends on how many pixels each letter gets.

Can you get text out of a video?

You can, by pausing on the frame you want and capturing that frame like any other screenshot. Any screen-capture OCR tool treats a paused video exactly the same as a static page.

Subtitles are a different problem with a better answer. If the video has real captions rather than burned-in text, YouTube will show a transcript directly, and that is far more accurate than reading the pixels.

Burned-in text in slides and lecture recordings is the case where frame capture genuinely wins. Accuracy depends mostly on resolution, so capture at full screen rather than from a small embedded player.