Copy text from scanned PDFs and images – offline
A scanned PDF, a photographed page, a screenshot, an error message in a dialog box: select the area and the text is in your clipboard. Recognition runs on your machine — the document is never uploaded.
Text recognition is included from Pro
Windows 10 and later · No admin rights needed · No account
Text you can see but cannot touch
To a computer, a scanned document is an image. The text in it is not writing but a pattern of pixels. You cannot select it, you cannot search it, you cannot copy it.
The same goes for screenshots, for error messages whose wording cannot be copied out, and for photographed pages.
The alternative is retyping. Annoying for three lines, working hours for three pages.
How you get the text out
Ctrl + Alt + T starts text recognition from any application. The hotkey is freely assignable.
A frame over the part of the screen — across several monitors too, at any resolution.
The recognised text is in your clipboard straight away. Ctrl + V, carry on working.
Image preparation — the actual difference
Poor originals are the norm: scanned crooked, grey, too dark, noisy, low resolution. Celerando prepares the image before recognition starts — which is where tools that pass the image through unchanged fall down.
Screenshot for screenshots and code, Photo / camera for photographed pages with uneven light, Scan / document for PDF pages and paper. Each profile enables exactly the stages that help that material.
In the manual Pro profile you control scaling, greyscale, contrast, binarisation, sharpening and noise reduction yourself — including the thresholds.
Tiny type is enlarged before recognition: 1.5× in the Screenshot and Scan / document profiles, up to 3× in the Pro profile. Without that step every engine returns gibberish on small text.
Why the profile matters: binarisation helps a clean scan and destroys half the letters in a photo with shadows. Sharpening helps the photo and creates artefacts in a screenshot. Hence three presets instead of one “enhance” button.
14 languages, mixed in the same document too
Celerando recognises with Tesseract 5 (LSTM). Eight languages ship with the installation, six more you download in the settings when you need them:
German, English, French, Spanish, Italian, Dutch, Polish, Portuguese.
Danish, Swedish, Finnish, Norwegian, Greek, Russian — a model of a few megabytes per language.
A German-English contract is recognised in one pass; the default selection is German + English.
- German
- English
- French
- Spanish
- Italian
- Dutch
- Polish
- Portuguese
- Danish
- Swedish
- Finnish
- Norwegian
- Greek
- Russian
The language models live in your user profile, not in the installation directory — so downloaded languages survive every app update.
What else comes with it
Every image screenshot lands in the editor: pen, marker, eraser, arrow, rectangle, text and colour picker.
Ctrl + Alt + S for the image screenshot, Ctrl + Alt + T for text recognition — both freely assignable.
After each screenshot Windows shows the result with a preview and buttons to edit or copy.
The document stays with you
Online text recognition means uploading your document. With a scanned contract, a personnel file or a medical letter that is not a small thing.
Celerando recognises on your machine. Neither the image nor the recognised text leaves it. What the app contacts at all: the download of each model and additional language you select, the optional CUDA runtime, the update check, license activation including the query of occupied seats, a single installation ping on the very first launch, and the addresses you enter yourself (web app tiles load the site's icon, the connections tile checks your servers). The ping carries a random installation ID, a truncated hash of the device fingerprint, and the app, .NET and Windows versions — no account, no name, no content. License activation and the seat query transmit the computer's hardware fingerprint, which is stored with the activation code. The IP address is not stored in plain text: a hash of it is deleted after 48 hours, while the approximate location derived from it (country, region, city) stays with the record for 24 months.
With the eight bundled languages, text recognition works without any connection at all — try it in flight mode.
Celerando, PowerToys and the Snipping Tool compared
Windows 11 now ships screen text recognition itself, and PowerToys has had it for longer. Both work locally — the difference is not privacy, it is what comes out on difficult originals.
| Celerando | PowerToys Text Extractor | Snipping Tool | |
|---|---|---|---|
| Where recognition happens | On your PC | On your PC | On your PC |
| Image preparation | Six stages, three profiles, upscaling | — | — |
| Languages | 14, several at once | Windows language packs | Windows language packs |
| Multi-monitor | Yes, across screens too | Per screen | Per screen |
| Screenshot editor | Pen, marker, arrow, text, picker | — | Yes |
| Price | Part of Pro, €36 once | Free | Included in Windows 11 |
As of September 2026. For a clean screenshot the built-in tool does the job — with a crookedly scanned contract, a photo with shadows or tiny type, the preparation decides the result.
Included in Pro
Text recognition is part of Celerando Pro — €36 once. Together with the Personal Dictionary, the full QuickShare and extended dictation.
You can try the free version at your leisure first.
Guides on this topic
- Convert an Image to Word: Editable Text from a JPG, Photo or ScanHow to convert a picture to text in Word: what Word can do on its own, where it stops, and how to get text from a JPG or photo into Word without uploading it.
- How to Edit a Screenshot on Windows: Crop, Highlight, Add TextCrop a screenshot, highlight something, add an arrow or text: what the built-in tools do on Windows 10 and 11, where they stop, and a faster way.
- How to screenshot one monitor on Windows 11 and 10Two monitors, but the screenshot should show only one? What each Windows screenshot key captures and how to grab exactly one screen.
- Can't copy text from a PDF? Four causes and what helpsCan't copy text from a PDF? How to tell a scan, a broken font, copy protection and a viewer problem apart, and what fixes each one, without uploading the file.
Frequently asked questions about text recognition
Does text recognition work offline?
Yes. With the eight bundled languages it needs no connection — the image and the recognised text stay on your machine. Only when you enable one of the six additional languages is its model downloaded once.
Which languages are recognised?
Fourteen: German, English, French, Spanish, Italian, Dutch, Polish and Portuguese are included right away; Danish, Swedish, Finnish, Norwegian, Greek and Russian are downloaded when you need them. Several languages can be active at once, so a German-English document is recognised in a single pass.
Does it work with badly scanned originals?
That is what the image preparation is for: scaling, greyscale, contrast, binarisation, sharpening and noise reduction, bundled into the Screenshot, Photo / camera and Scan / document profiles. With very poor originals every recogniser hits its limits though — a page a human can barely decipher will not become clean text here either.
What can Celerando do that PowerToys or the Snipping Tool cannot?
The image preparation before recognition (three profiles, six individually controllable stages, upscaling, up to 3× in the Pro profile), free language selection with several languages at once, and selection across monitors. For a clean screenshot in your system language the built-in tools are perfectly sufficient — the differences show up with scans, photos and small type.
Which engine is behind it?
Tesseract 5 with LSTM recognition, an established open-source project. The language models are the official Tesseract data; Celerando provides the image preparation, the area selection and the integration into Windows.
Does it work with several monitors?
Yes, across monitors too and at any resolution — a frame may cross the screen boundary.
Can I pull text out of a PDF without opening it?
Text recognition works on what is visible on screen, so the PDF has to be open — then you drag the frame across the page. For scanned PDFs that is exactly the route, because their text cannot be selected anyway.
Where are the language models stored?
In your Windows user profile under %LOCALAPPDATA%\Menterium\Celerando, not in the installation directory. That is deliberate: an app update replaces the installation directory, and a language downloaded there would vanish without a trace.
Which edition includes text recognition?
From Pro. The free version does not include it — but the Dashboard, voice input in your system language and QuickShare for text can be tried without a time limit.
Do I need admin rights?
No. Celerando installs into your user profile, and so do the language models.