How to Extract Text from an Image or Screenshot on Windows

There is text right there on your screen — in a screenshot, a photo of a whiteboard, a scanned document, a recipe someone texted you as a picture. You can read it with your eyes. But when you try to copy it, nothing happens. The text is trapped inside the image, and Windows treats the whole thing as a single block of pixels rather than readable words.

This is one of those everyday frustrations that feels like it should have been solved by now. You can ask your phone to identify a plant from a photo, but copying a phone number from a screenshot still requires either retyping it manually or finding the right tool.

The technology that solves this is called OCR — Optical Character Recognition. It has been around for decades, but it has gotten dramatically better in the last few years. The question is not whether it works (it does) but which approach makes sense for what you are trying to do.

Quick answer: Windows has a few built-in ways to extract text from images, but they are limited and not always obvious. For reliable text extraction from screenshots, photos, and scanned documents, a dedicated OCR tool like OCR Text Recognition Tool from the Microsoft Store handles the job with minimal setup — open the image, extract the text, copy it. For occasional use, Windows PowerToys also includes a text extraction feature worth knowing about.

What OCR actually does

OCR is the process of analyzing an image to identify the characters in it and converting them into text you can select, copy, and edit. At a high level, the software looks at the shapes in the image, compares them against known character patterns, and produces its best guess at what each character is.

Modern OCR goes well beyond simple pattern matching. Current engines use machine learning models that understand context — they know that “rn” in a certain font looks almost identical to “m,” and they use the surrounding words to figure out which one it actually is. They can handle skewed text, uneven lighting, handwritten characters (to varying degrees), and multiple languages on the same page.

The accuracy depends on a few factors: how clear the source image is, what font the text uses, whether the text is printed or handwritten, and how much visual noise is in the image. A clean screenshot of a document converts almost perfectly. A blurry photo of a crumpled receipt taken in bad lighting is a harder problem.

📊 OCR methods on Windows compared

The situations where you need this

Before diving into tools, it helps to understand the common scenarios. People search for text extraction for surprisingly different reasons, and the best approach depends on which one you are dealing with.

Screenshots of error messages or code. A coworker sends you a screenshot of a bug. You need to search for the error text or paste it into a ticket. Retyping a stack trace character by character is tedious and error-prone. OCR extracts the exact text in seconds.

Scanned documents. You have a scanned contract, a receipt photo, or a PDF that is actually just images of each page. The text looks normal on screen, but nothing is selectable. OCR converts the image into actual text you can copy and search.

Photos of printed material. A page from a book, a whiteboard after a meeting, a nutrition label, a street sign in another language. Any time you can photograph text but cannot digitally select it, OCR is the bridge.

Data locked in images. Someone sends a table as a screenshot instead of a spreadsheet. A chart has labels you need to reference. A presentation slide has text you want to quote. All of these require extracting the text from the image rather than the underlying file.

Text in another language. You have an image with text in a language you do not read. Extracting the text first and then translating it is more reliable than trying to translate directly from the image, because translation tools work better with clean text input.

Method 1: PowerToys Text Extractor

Microsoft PowerToys is a free collection of utilities for Windows, and one of them — Text Extractor — does basic OCR. If you already have PowerToys installed, you can use it immediately. Press Win + Shift + T and draw a box around the text you want to extract. The recognized text goes straight to your clipboard.

It works well for clean, well-lit text in standard languages. Where it falls short is on complex images, multi-language content, or situations where you need more control over the process. There is no preview of what it recognized, no ability to correct errors before copying, and the language support depends on which OCR language packs you have installed on Windows.

For quick, occasional use — grabbing a line of text from a screenshot, copying an address from an image — it is surprisingly useful. For anything more involved, it feels limited.

If you do not have PowerToys installed, it is available from the Microsoft Store or GitHub. The install is straightforward, though you get the entire PowerToys suite, not just the OCR feature.

Method 2: A dedicated OCR app

If text extraction is something you do more than occasionally, a dedicated OCR app is the more practical choice. The workflow is simpler: open your image, click extract, and the recognized text appears ready to copy. No keyboard shortcuts to memorize, no drawing selection boxes on screen, no guessing whether the extraction worked until you paste it somewhere.

Dedicated apps also tend to handle edge cases better — images with mixed languages, text at odd angles, lower-quality photos, scanned documents with background noise. They invest their engineering into the OCR pipeline specifically, which shows in the accuracy on difficult inputs.

Some OCR apps process images using cloud-based recognition engines. This is a practical design choice: cloud engines are trained on vastly larger datasets and tend to be more accurate, especially for handwriting, unusual fonts, and non-Latin scripts. The trade-off is that the image leaves your machine briefly for processing. For most use cases — extracting text from a screenshot, reading a scanned receipt — this is not a concern. If you are working with classified material, it is worth knowing.

Method 3: OneNote’s hidden OCR feature

OneNote has OCR built in, but it is not obvious. If you paste or insert an image into a OneNote page, right-click it and select “Copy Text from Picture.” OneNote processes the image and copies the recognized text to your clipboard.

The accuracy is decent for clean images and printed text. The downside is the workflow: you have to open OneNote, create or navigate to a page, insert the image, wait a moment for processing, then right-click. If you already use OneNote for other things, this is a neat trick. If you do not, opening a note-taking app just to extract text from an image is an awkward detour.

Another quirk: OneNote sometimes needs a few seconds to process the image before the “Copy Text” option appears. If you right-click immediately after inserting, the option may not be there yet.

Method 4: Google Lens in the browser

Google Lens can extract text from images directly in Chrome. Right-click any image on a web page and select “Search image with Google Lens,” then switch to the “Text” mode. You can select and copy the recognized text.

For images already in your browser — a screenshot posted on a website, a document preview, an embedded image — this is convenient because there is nothing to install. For images on your desktop or in a folder, you would need to drag them into the browser first, which adds a step.

The OCR quality is strong, especially for multiple languages. Google’s recognition engine is one of the best available, and Lens benefits from the same technology. The limitation is that it only works inside Chrome and requires an internet connection.

What affects OCR accuracy

Not all images produce the same results, regardless of which tool you use. Understanding the factors helps you set expectations and, when possible, improve the source image before extracting.

Resolution matters. Higher-resolution images produce better OCR results. A 300 DPI scan converts almost perfectly. A 72 DPI screenshot of a small text block may produce errors. If you have control over the source, capture at the highest resolution you can.

Contrast matters more than color. OCR engines convert images to high-contrast internally before processing. Black text on white background produces the best results. Light gray text on a slightly lighter gray background — the kind of thing that looks fine to your eyes — can confuse the engine because the contrast ratio is too low.

Straight text is easier than skewed text. If you are photographing a page, try to capture it straight-on rather than at an angle. Modern OCR can handle some skew, but every degree of rotation reduces accuracy slightly. If you are scanning, use a flatbed scanner rather than a phone camera for the cleanest results.

Printed text versus handwriting. Printed text in standard fonts converts very accurately — 99%+ for clean images. Handwriting recognition has improved dramatically but still varies. Neat, consistent handwriting works reasonably well. Messy handwriting or unusual styles remain difficult for any OCR engine.

Noise and artifacts. Smudges, coffee stains, crease lines, and compression artifacts all reduce accuracy. JPEG compression in particular can blur character edges just enough to cause misreads. If the source is a heavily compressed image, the OCR engine has less to work with.

Dealing with scanned PDFs

A specific variant of the text extraction problem is the scanned PDF. This is a PDF where each page is actually a photograph — the text looks normal when you view it, but try to select it and nothing highlights. The PDF viewer treats each page as a single image.

To check whether a PDF is scanned or text-based, open it and try to click-and-drag to select a word. If you can highlight individual words, it is text-based and you do not need OCR — you can copy the text directly. If nothing highlights, or the entire page selects as one block, it is scanned.

For scanned PDFs, the approach is the same as for any image: run OCR on it. Some OCR tools accept PDF files directly and process each page as an image. Others require you to convert the PDF pages to images first. Either way, the result is extractable text.

If you also need the document back in PDF format with the text embedded (so others can search and select it), you need a tool that can create a “searchable PDF” — an OCR’d layer placed behind the image so the visual appearance is preserved while the text becomes selectable. Adobe Acrobat does this well. Some dedicated OCR tools offer it too.

Troubleshooting

Extracted text has a lot of errors. The source image is likely low resolution, low contrast, or contains unusual fonts. Try improving the image first — increase contrast, crop to just the text area, or rescan at higher resolution.

Text comes out in the wrong order. Multi-column layouts and complex page designs can confuse OCR engines about reading order. If the text is extracted but the paragraphs are scrambled, you may need to extract smaller sections at a time — one column, one block, one paragraph.

Special characters or symbols are wrong. OCR engines handle standard alphabetic text best. Mathematical symbols, currency signs, and special punctuation are harder to recognize and may appear as wrong characters. Proofread these carefully.

Non-Latin scripts have lower accuracy. OCR for Chinese, Japanese, Korean, Arabic, and other non-Latin scripts has improved but is generally less accurate than Latin-script recognition. Make sure your OCR tool supports the specific language and that any language packs are installed.

Handwriting is not recognized at all. Not every OCR engine supports handwriting. Those that do typically require the handwriting to be fairly neat and consistent. If you need to digitize handwritten notes regularly, look for a tool that specifically advertises handwriting recognition.

FAQ

Can Windows extract text from images natively?

Not through any obvious built-in feature. PowerToys Text Extractor adds this capability if installed. OneNote has a hidden OCR feature. Neither is a one-click solution out of the box.

Is OCR accurate enough to trust without proofreading?

For clean, high-resolution images with printed text, accuracy is typically 99%+. For lower-quality images, handwriting, or unusual fonts, always proofread. Never assume OCR output is perfect on important documents.

Can I extract text from a photo taken with my phone?

Yes. The accuracy depends on the photo quality. Good lighting, steady focus, and a straight-on angle produce the best results. Blurry or angled photos will have more errors.

Does OCR work with handwriting?

To a degree. Neat, consistent handwriting in common scripts can be recognized with reasonable accuracy. Messy handwriting, unusual scripts, or highly stylized writing remains difficult for all OCR tools.

Can I extract text from a video or live screen?

Not directly. You would need to take a screenshot of the frame with the text you want, then run OCR on that screenshot. Some screen capture tools let you freeze a frame for this purpose.

What about extracting text in multiple languages?

Most modern OCR tools support dozens of languages. If the image contains text in more than one language, some tools handle this automatically while others need you to specify the languages in advance. Multi-language support is one area where dedicated OCR tools tend to outperform built-in options.

Sources

• Microsoft PowerToys: learn.microsoft.com/en-us/windows/powertoys/text-extractor
• Google Lens: lens.google.com
• OneNote OCR: support.microsoft.com/en-us/office/copy-text-from-pictures
• Wikipedia — Optical Character Recognition: en.wikipedia.org/wiki/Optical_character_recognition

Final takeaway

Extracting text from images is a solved problem — the tools exist, the accuracy is high, and the process is fast. The real question is just which approach fits how often you need it and how smooth you want the experience to be. PowerToys works for occasional quick grabs. OneNote’s hidden feature works if you are already there. For anything more regular, OCR Text Recognition Tool gives you 5 free extractions per day with multi-language support and a straightforward workflow — open the image, extract, copy.

Whatever you choose, stop retyping text from screenshots. Life is too short for that.