PowerToys Text Extractor: What It Copies From a Screen
By Schema LawPublished Updated
PowerToys Text Extractor copies the words inside a rectangle you drag on your Windows screen and puts them on the clipboard. It reads whatever is showing, including text inside images and videos. Microsoft's own page for the utility now recommends the Snipping Tool for this job, and warns that the copied text needs proofreading.
The source for every statement about the utility here is Microsoft Learn's page on the PowerToys Text Extractor utility, which was last updated in August 2025 when this post was written. PowerToys is a set of utilities that Microsoft describes as being "for power users", installed separately from Windows.
What does PowerToys Text Extractor do?
PowerToys Text Extractor copies text from a region of the screen to the clipboard. Microsoft's page says the utility "enables you to copy text from anywhere on your Windows screen, including inside images or videos", and that "the text will be saved to your clipboard." It uses OCR, and the page credits the code to Joe Finney's Text Grab.
That description fixes what the tool is and what it is not.
- Input: the screen. Text Extractor reads pixels that are currently visible. It does not open a PDF, a TIFF or a folder of scans.
- Output: the clipboard. Nothing is saved. There is no file, no page image and no record of where the words came from. Paste before you copy anything else.
- Use: a snippet. A case number from a docket screenshot, a paragraph from a scanned exhibit open in a viewer, a line of text burned into a video frame, the wording of an error message you need to send to IT.
The page does not publish a list of fonts, sizes or image qualities the utility can read, and it makes no accuracy claim. What it does say about quality is one warning: the produced text "may not be perfect".
The Text Extractor shortcut and how the capture overlay behaves
The default shortcut is the Windows key + Shift + T. Microsoft says pressing it shows "an overlay on the screen", after which you "click and hold your primary mouse button and drag to activate your capture." Capture mode closes as soon as the text in the region has been recognized and copied. Esc cancels at any moment.
| Action | Keys or mouse, per Microsoft |
|---|---|
| Start a capture | Windows key + Shift + T (the default; it can be changed) |
| Choose the region | Hold the primary mouse button and drag |
| Move the region without resizing it | Hold Shift while dragging; release Shift to resize again |
| Cancel | Esc |
| Finish | Release the mouse; the mode closes after the text is copied |
Two settings are documented, both in the PowerToys Settings menu: the activation shortcut, and a preferred language for recognition.
Microsoft's page describes no preview of the result, so the way to see what was recognized is to paste it. Paste into a plain text editor first. You will see the raw characters and line breaks before a word processor restyles them, which makes a dropped line or a broken word easier to spot.
A habit that helps: before you press the shortcut, zoom the document so the passage is large on screen, and scroll so the whole passage is visible at once. The utility can only read what is showing.
Which languages can PowerToys Text Extractor read?
Only languages whose Windows OCR language pack is installed on that computer. Microsoft's page states that Text Extractor "can only recognize languages that have the OCR language pack installed", and that the default language is based on your Windows system language and keyboard settings.
Microsoft gives PowerShell commands for finding out what you have. To list the languages the OCR engine can use right now, the page says to run these in Windows PowerShell, and specifically not PowerShell 7:
[Windows.Media.Ocr.OcrEngine, Windows.Foundation, ContentType = WindowsRuntime]
[Windows.Media.Ocr.OcrEngine]::AvailableRecognizerLanguagesTo list every OCR language pack Windows offers and whether each is installed, Microsoft says to open PowerShell as an Administrator and run:
Get-WindowsCapability -Online | Where-Object { $_.Name -Like 'Language.OCR*' }Each result has a name such as Language.OCR~~~es-MX~0.0.1.0 and a state of Installed or NotPresent. Microsoft's page explains that a pack marked NotPresent must be installed before the language can be recognized, and that a language missing from the output altogether is not supported by OCR. The same page gives the install command pattern for a pack and notes that installing one "may take several minutes to complete".
For a law office this matters in a specific case: a Spanish-language birth certificate or a French contract on a machine that has only the English pack. On a firm-managed computer you may not have administrator rights, so the request goes to IT. If the utility shows the message "No Possible OCR languages are installed", Microsoft says that means no recognition language is available at all.
Why Microsoft recommends the Snipping Tool over Text Extractor
Microsoft's page carries a note near the top: "It's recommended to use the Snipping Tool instead of the Text Extractor for capturing screenshots." The page does not give a reason. It does not say Text Extractor is being removed, and it does not compare the two tools' results.
So the honest reading is narrow. The vendor that publishes both tools points you to the one built into Windows, and says nothing more. What you can compare for yourself, from the two documentation pages:
| PowerToys Text Extractor | Snipping Tool | |
|---|---|---|
| How you get it | Installed as part of PowerToys | Part of Windows |
| What you keep | Text on the clipboard only | A snip you can view, plus text you copy from it |
| Text commands documented | Copy the region's text | Select and copy, Copy all text, Quick redact |
| Languages | Installed OCR language packs, with a preferred language setting | Not addressed on the support page |
For a firm, the first row can decide it. A utility that needs a separate install needs someone to approve it and keep it updated. The post on Snipping Tool OCR and its Text actions button covers what the built-in tool does, including Microsoft's statement about where it processes text.
If Text Extractor is already installed and you like the single shortcut, Microsoft's note is a recommendation and not a withdrawal. Test both on the kind of page you copy from most and keep the one whose output you correct least.
Proofreading what Text Extractor puts on the clipboard before it reaches a brief
Microsoft tells you to proofread. The page's own warning, marked Important, reads: "The produced text may not be perfect, so you have to do a quick proof read of the output." In legal writing, a quick read is not enough for a quotation. Compare the pasted words with the source, character by character where it counts.
A check that fits a short passage:
- Numbers first. Dates, dollar amounts, section and paragraph numbers, case numbers, Bates numbers. Read each digit against the screen.
- Names and defined terms. A party's name with one wrong letter looks like a different party.
- Small words that flip meaning. "Not", "nor", "unless", "shall" and "may". Confirm none was dropped at a line break.
- Line ends. A word hyphenated across two lines can paste as two fragments or lose its hyphen.
- Punctuation inside the quotation marks. Commas and periods are small on screen and easy for recognition to miss.
- The citation. The clipboard carries no page number. Write the document, page and line next to the paste before you close the viewer.
The post on why OCR output is full of errors sets out a fuller method for checking recognized text against the page.
Remember what you have when you are done: a few correct sentences in a draft. The scanned document they came from is still an image, and nobody can search it. The Snipping Tool post explains clipboard text versus a searchable file in full. When you need the whole document as text, a screen utility is the wrong tool; a converter such as Schema Law reads an uploaded PDF, PNG or JPEG into text and tables that you export, and the how it works section of the homepage shows the steps.
Paste into a plain note and read it against the screen for the party names, a date on the notice, and a line from the pleading or the contract. If the rectangle covered the signature block of an affidavit, read that line a second time. The rest of the page stays on the scan, where you can still see it.