Local or upload: where your document goes
OCR tools differ most in where the recognition happens. Engines and desktop apps read the file on your machine; web services receive the upload and delete it after a while. For contracts, IDs and anything with personal data, the first group is the safer choice:
| Where it runs | Tools | What the official page says |
|---|---|---|
| On your computer | Tesseract, OCRmyPDF, NAPS2, gImageReader, EasyOCR, NormCap, PowerToys Text Extractor, Text Grab | NormCap: detection is performed on your device; Text Grab: all OCR runs entirely on your device |
| In the browser, nothing sent | Scribe OCR | The entire program runs in your browser, no data is sent to a remote server |
| Uploaded to a server | OnlineOCR, OCR.space, i2OCR, Pen to Print, Transkribus, Adobe Scan | OCR.space: files and text deleted immediately after processing; i2OCR: deleted after 30 minutes; Transkribus: hosted in Austria |
Our image to text and OCR PDF tools belong in the second row: they run the recognition in your browser. The PDF to text tool extracts an existing text layer without OCR, and scan to PDF turns photos of pages into a PDF first.
Free limits of the web tools
The online services are the quickest for a single page, with these caps from their own pages. OnlineOCR lets a guest convert 5 files per hour with a file size of up to 15 MB, outputs TXT, DOCX and XLSX in 46 recognition languages; registered users get 200 MB files, 50 free pages, all pages of a multi-page PDF and ZIP uploads. OCR.space is free to use with no registration, takes PNG, JPG, WebP and PDF up to 5 MB, supports multi-page documents and lists over 200 additional languages for its engine 3. i2OCR is free for one image or PDF page at a time in 128 languages, with plain text, Word, HTML or searchable PDF as output; bulk OCR needs a premium plan.
For whole books or stacks of scans, the free caps run out fast. That is where the local tools below have no limit at all.
Open source engines and the apps built on them
Tesseract is the engine behind most free OCR: an Apache-2.0 library and command line tool that recognizes more than 100 languages and writes plain text, hOCR, PDF, TSV, ALTO and PAGE. It has no GUI of its own, so the apps around it matter:
- OCRmyPDF (MPL-2.0) adds an OCR text layer to scanned PDF files and produces PDF/A, on Linux, Windows, macOS and FreeBSD.
- NAPS2 (GPL-2.0 or later) scans from WIA, TWAIN, SANE and ESCL scanners and runs OCR in any of over 100 languages, free for individuals and businesses with no ads.
- gImageReader (GPL-3.0) is a Gtk/Qt front end that imports PDFs and images from disk, scanners, clipboard and screenshots and outputs text, hOCR or PDF.
- Scribe OCR (AGPL-3.0) runs Tesseract models in the browser with a proofread mode.
- NormCap (GPLv3) and Text Grab (MIT) use Tesseract and, on Windows, the built-in OCR to read a region of the screen.
EasyOCR (Apache-2.0) is the other engine in the list: a Python package with 80+ supported languages covering Latin, Chinese, Arabic, Devanagari and Cyrillic scripts, on CPU or GPU.
Reading text off the screen
Screen OCR tools skip the screenshot: press a shortcut, draw a rectangle, and the text lands in the clipboard. PowerToys Text Extractor opens with Win+Shift+T, uses the OCR language packs installed in Windows and is part of the MIT-licensed PowerToys. Text Grab, which the Microsoft docs name as the basis of that utility, adds Tesseract and the WinAI models on Copilot+ PCs. NormCap does the same on Windows, macOS and Linux, also reads QR codes and barcodes, and detects email addresses and URLs.
Expect to proofread: the Microsoft page itself says the produced text may not be perfect. Numbers, columns and small type are the usual trouble spots.
Handwriting and phone cameras
Handwriting needs different models. Pen to Print converts neat, cursive or messy handwriting from scans, images or PDFs online, on iOS and Android, and on Windows and Mac, with 10 free pages to try and paid plans after that. Transkribus, built for archives and researchers, gives 50 free credits per month, enough for around 50 pages, with handwritten text recognition at 1 credit per page and a Scholar plan at 99 euros a year.
On a phone, Adobe Scan is free to download on iOS 15 and Android 8.0 or later and turns scans into searchable PDFs with OCR. If you already have the PDFs, our list of free PDF editors covers what to do with them next, and for spoken rather than written words the free transcription tools are the right list.
Frequently asked questions
Which OCR tools keep my files on my computer?
Tesseract, OCRmyPDF, NAPS2, gImageReader and EasyOCR run on your machine. NormCap states that detection is performed on your device, and Text Grab that all OCR runs entirely on your device. Scribe OCR runs entirely in your browser and sends no data to a remote server.
What are the free limits of online OCR?
OnlineOCR allows a guest 5 files per hour up to 15 MB each, and gives registered users 200 MB files and 50 free pages. OCR.space is free with no registration up to 5 MB per document. i2OCR is free for one image or PDF page at a time, and bulk OCR needs a premium plan.
Which free tool makes a scanned PDF searchable?
OCRmyPDF adds an OCR text layer to scanned PDFs and outputs PDF/A, using Tesseract for more than 100 languages. gImageReader imports PDFs and can generate a PDF from its hOCR output. NAPS2 scans directly to a PDF with OCR, and Adobe Scan does the same on a phone. Our OCR PDF tool runs in the browser.
Can free OCR read handwriting?
Pen to Print is built for handwriting, including cursive, and offers 10 free pages before a paid plan. Transkribus includes 50 free credits per month for handwritten text recognition at 1 credit per page. The general engines are made for printed text; i2OCR notes that handwriting may reduce precision.
How many languages do the engines support?
Tesseract recognizes more than 100 languages, and OCRmyPDF and NAPS2 inherit that through its language packs. EasyOCR lists 80+ languages. Among the web tools, i2OCR states 128 languages, OnlineOCR 46, and OCR.space over 200 additional languages for its engine 3.
Is there a free way to copy text from the screen?
PowerToys Text Extractor (Windows, MIT) copies text from any screen region with Win+Shift+T using the installed OCR language packs. Text Grab (Windows, MIT) and NormCap (Windows, macOS, Linux, GPLv3) do the same with Tesseract, and NormCap also reads QR codes and barcodes.