OCR

This article is the OCR job. Words are on the screen and you cannot select them. A screenshot, a paused video, a PDF page in a viewer, a photo opened in Photos, a dialog that blocks copy. You want those pixels turned into text on this PC. Taking words from a picture file you already saved sits on Extract text from image. The short install sits home.

This is Text-Grab. MIT. Joseph Finney publishes it. We did not make it. Official files live on GitHub Releases. The repo is TheJoeFin/Text-Grab. This site does not host the installer. Skip Softonic, CNET, and unofficial wrappers.

The button on this page goes to Install on the home page. Trust notes sit on Image to text download safe.

Install so you can use OCR on Windows

There is no winget command. Do not paste a made-up package id. Copy this filename:

Text-Grab.exe

Open the latest GitHub release. For most PCs take Text-Grab.exe. On ARM Windows take Text-Grab-arm64.exe. Self-contained zips also sit on that release if you want a zip. Live tag observed: V4.15.0, 1 August 2026. Do not invent a newer number here. File notes sit on Image to text releases. LICENSE sits on Image to text GitHub.

Run the exe on the same machine. Files stay on GitHub. If SmartScreen warns on a first run, read the publisher. Prefer the GitHub asset over a renamed copy from search.

How to use OCR

1. Open GitHub Releases, pick Text-Grab.exe for x64 or Text-Grab-arm64.exe for ARM, and run it. There is no winget command.
2. Open Text-Grab. Choose Full-Screen Mode. You should get a grab on this desktop, not a browser tab that uploads a photo.
3. Draw a box around the words on this PC. Click once if you only need a single word. Windows draws a box around each recognized word.
4. The recognized text goes to the clipboard. Recognition runs on this device. Nothing here sends the screen to a website.
5. Paste where you need it, or open Edit Text Window in the app to clean the lines. Press Escape to cancel. Right-click Cancel also exits. Alt+F4 exits too.

That list is the OCR path. You are reading pixels that are already on this PC. You are not opening a saved image as the whole job. That file job sits on Extract text from image.

What OCR is on this PC

OCR is optical character recognition. The program looks at a picture of letters and writes real text you can paste. Windows 10 and later ship an OCR API. Text-Grab is one MIT program that uses that API in a desktop window. You install it, you open Full-Screen Mode, you draw a box, you paste.

This page stays on recognition of what you can see right now. A file in Downloads is a different article. If the words only exist as a png you already stored, open that job instead. Stay here when the words are on screen and copy is blocked.

The app does not create the words. If the screen is a solid color, a closed lid, or a window you minimized, there is nothing to read. Show the words. Then grab.

Full-Screen Mode

Full-Screen Mode is the usual OCR start. It behaves like a screenshot tool, except the output is text, not a png. Draw a rectangle around the line you need. The engine reads that region. The clipboard gets the result.

Click once to try a single word. That works because the Windows OCR API puts a bounding box around each recognized word. If you click or select an area with no text, the window stays active so you can try again. Empty grab is not a broken install. Move the box. Try the line that actually has letters.

To exit, press Escape. Right-click and choose Cancel. Or press Alt+F4. Closing the grab is enough for one pass. You do not reinstall for the next region.

Start with one line you can check. A heading, a serial, a sentence you already know. Paste into Notepad. Confirm the letters match. Then grab the next region. An OCR session that has never pasted one good line is a bad place to sweep a whole desktop of tiny type.

Grab Frame when the box should stay

Grab Frame is a movable OCR window you keep over part of the screen. Position the frame over the text. Grab by searching in the frame, clicking a word border, or clicking Grab. Same engine as Full-Screen Mode. Same strengths. Same misses.

Use Grab Frame when the same corner of an app keeps showing values you need, and drawing a new full-screen box every time is noisy. Resize the frame. Accuracy often improves when the frame is tight on the letters and not on a busy wallpaper.

We will not invent extra menu names. Use the controls the app actually shows. If the frame is covering the wrong monitor, move it. If the text is too small, enlarge the source window first, then grab again.

First run

After you run the exe, open the app once before you hunt a second program named OCR. Pick Full-Screen Mode if the first-run choice asks. Grab one line you trust. Confirm paste works. If the clipboard is empty, you have not selected a region yet, or the region had no letters, or the path failed.

Windows may show a SmartScreen prompt on a first exe. Read the publisher. Prefer GitHub Releases. The exe path is the honest one on this site. Do not fetch a second setup from search.

Updates live on GitHub Releases. We do not invent a newer number than V4.15.0 here. Read that page. Usual Windows pick is still Text-Grab.exe. The arm64 exe is still for ARM Windows. The self-contained zip is still a zip of this project.

When recognition copies nothing useful

A region of icons will not become a paragraph. A blurry photo of a page, zoomed until the letters smear, will miss. A screenshot of a screenshot stacked three times can fail. Show a sharp line. Grab that.

Handwriting is harder than printed UI type. Stylized fonts, outlined titles, and text that is a texture in a game can miss. Try a tighter box. Try Grab Frame. If the engine still writes garbage, the pixels are the limit. This page will not repair a photo. It will not host a second engine as a download from us.

Windows OCR languages follow language packs in Windows Settings. If you grab a script Windows has not added, you can get empty output or the wrong letters. Add the language pack on the PC, then try again. We do not invent a language list here. Check Windows.

A paused video frame is still pixels. OCR can read a subtitle or a slide if the frame is sharp. A moving frame is a bad grab. Pause. Then draw the box.

Edit Text Window after a grab

Full-Screen Mode is for speed. Edit Text Window is for cleanup. Paste or send the grab there when the clipboard dump has extra spaces, stacked lines, or mixed case you do not want. Make a single line. Trim empty lines. Toggle case. Find and replace. That is still OCR work: you recognized the words, now you make them usable.

If the grab was a table on screen, Spreadsheet mode in the app can keep columns after recognition. We will not invent extra column names. If the preview looks like stacked cells, try that mode. If it looks like a sentence, stay in plain text and paste.

You can skip Edit Text Window. Many OCR jobs end at Ctrl+V in the app you already had open. Cleanup is optional. Recognition is the job on this page.

School PCs, labs, and uninstall

A lab image may block unknown exe files. Then this OCR install is closed for that hour. Do not fetch a wrapper to get around policy. Use what is already on the image, or wait until you control the PC.

A borrowed laptop you return in twenty minutes is a bad place to leave Text-Grab.exe. Use what is already on the machine, or do not install.

Uninstall is Apps in Windows if the build registered there, or delete the unpacked zip if you used a zip. Quitting the window is not uninstall. If you only wanted one grab, quit first. If you want the program gone, remove it. Fetch again from GitHub Releases, not from a wrapper.

x64 exe, arm64 exe, and zip

Usual Windows pick is Text-Grab.exe. ARM Windows uses Text-Grab-arm64.exe. The same GitHub release also lists self-contained zips. They are different packages of the same project. Pick one path. Do not run the exe and unpack the zip on top of each other because the first one felt slow.

Match the file name on the release. Do not invent a newer tag than V4.15.0. Hash notes, if any, live on GitHub Releases, not here. The install pointer is Install on the home page.

What this OCR job will not do

It will not be the saved-picture job. That sits on Extract text from image. It will not upload a screenshot to a website. It will not host the installer.

It will not be a scanner. It will not add a printer. It will not be a star rating. It will not invent a winget id. Files stay on GitHub.

It will not fix a photo that never had readable letters. It will not wait on a wrapper named setup. Skip Softonic, CNET, and unofficial wrappers.

Shared machines and the next grab

On a family PC, say if you installed a new OCR program. On a work PC, follow local policy. If policy blocks unknown exe files, stop. Softonic is still the wrong host in that hour.

Grab the next region the same way. Install once. Recognize many times. You do not reinstall for each box. If a new region fails, the pixels are the suspect, not a missing second OCR app.

How do I use OCR?

Take Text-Grab.exe from GitHub Releases, or the arm64 exe. Run it. Open Full-Screen Mode. Draw a box around the words. Paste. There is no winget command. The install pointer is Install on the home page.

Does OCR need the internet?

Recognition in this program runs on the device. You still need the network to fetch the exe from GitHub the first time. After the file is on this PC, a grab does not upload the screen. This site does not host the installer.

Why did OCR copy nothing?

The box may have missed the letters, the type may be too small or too stylized, or Windows may lack the language pack. Show a sharp line. Draw a tighter box. Paste once into Notepad to confirm. Empty clipboard after an empty region is expected. Try again.

Did you make this OCR?

No. Joseph Finney publishes Text-Grab under MIT. Official files live on GitHub Releases. This site only points at that page. We do not host the installer. Back home if you want the short version.