PDF OCR
A scanned PDF is just a stack of pictures — you cannot select, copy or search anything in it. This tool runs optical character recognition on every page directly in your browser and saves the result as a plain-text file. Pick one of six languages, drop your PDF, start the recognition and watch the per-page progress until the .txt download is ready.
100% in your browser. Your file never leaves your device.
Features
- Recognizes printed text in scanned PDFs entirely in your browser — no server upload.
- Six recognition languages to choose from: English, German, French, Spanish, Italian and Portuguese.
- Each page is rendered at double resolution so small print stays legible for the engine.
- Live progress bar with a page counter and a cancel button that stops the job.
- Output is a plain .txt file with a clear separator between the pages.
- Engine and language data load only when you start and are cached for later runs.
About this tool
Under the hood, two open-source libraries share the work. PDF.js, Mozilla's PDF renderer, opens your file and draws every page onto an invisible canvas at twice its normal size. Tesseract.js, a WebAssembly port of the Tesseract OCR engine, then analyses each of those page images, detects lines and characters, and returns the recognized text page by page.
Because the OCR engine is heavy — roughly three megabytes of code plus around ten megabytes of language data — it is loaded only at the moment you click Start text recognition, never while you are just reading the page. The language file is fetched once per language and cached by your browser, so a second run with the same language starts noticeably faster.
Your PDF itself never leaves your device: it is read into memory, rendered, recognized and discarded when you leave the page. The result is saved as a .txt file named after the original document, with a heading marking each page. If you cancel midway, the recognition worker is terminated immediately and nothing is stored anywhere.
How to OCR a PDF
-
Drop your scanned PDF
Drag the file onto the drop zone or click it to open the file picker. The tool reads the PDF locally and shows its name and size in a small card.
-
Choose the recognition language
Pick the dominant language of the document from the dropdown — English, German, French, Spanish, Italian or Portuguese. The first run per language downloads about ten megabytes of language data, which is then cached.
-
Start the text recognition
Click Start text recognition. The tool now loads PDF.js and the Tesseract engine on demand, opens your document and begins rendering the first page. Nothing is transferred except the public library files.
-
Wait or cancel anytime
A progress bar counts through the pages one by one — recognition runs sequentially, so long documents take a while. The Cancel button terminates the OCR worker instantly and returns you to the start.
-
Save the text file
When the last page is done, a .txt file named after your PDF is offered for download. Each page appears as its own section, so you can match text back to the original.
Typical use cases
-
Digitizing paper records
A folder of scanned invoices, contracts or meeting minutes is useless when you need one specific clause. Run each PDF through OCR and store the text output next to the scan — suddenly your archive is searchable with any desktop search tool, and you can copy amounts, dates and names instead of retyping them.
-
Quoting from scanned literature
Students and researchers often work with scanned book chapters or old journal articles where copying is impossible. OCR the chapter once and you can quote passages, feed the text into reference managers or translation tools, and search the whole document for a term instead of skimming page images by eye.
-
Confidential scans without cloud upload
Medical reports, payroll sheets or legal correspondence should not be handed to an anonymous OCR server just to get the text out. Because recognition here happens inside your own browser, the document never crosses the network at any point — a practical option when confidentiality rules, professional secrecy or plain caution rule out cloud services.
Tips & limitations
- The output is plain text only: columns, tables, font styles and images are lost, and multi-column layouts can come out in a jumbled reading order. Expect to reformat the result.
- Recognition runs entirely on your device, one page at a time. A hundred-page scan can take several minutes and noticeable CPU — keep the tab open and in the foreground until it finishes.
- Recognition quality depends heavily on the scan: aim for at least 200 DPI, straight pages and good contrast. Skewed, dark or heavily compressed phone photos produce many more errors.
- Only one language model is active per run. For a document that mixes languages, choose the dominant one and expect errors in the other — or run the tool twice with different settings.
Frequently asked questions
Is my scanned PDF uploaded to a server?
No. The PDF is opened and processed entirely in your browser. The only network traffic is the one-time download of the OCR engine and the selected language data, both of which are public library files that contain nothing of yours.
Why does the first run take so long to start?
Clicking Start text recognition triggers the download of the Tesseract engine (about three megabytes) plus roughly ten megabytes of language data. Your browser caches both, so later runs with the same language skip the download and start almost immediately.
Which languages can it recognize?
Six languages are built in: English, German, French, Spanish, Italian and Portuguese. Only one model runs at a time, so pick the language that dominates your document. Tesseract supports further languages, but this browser version does not offer them.
Can I get a searchable PDF with an invisible text layer?
Not in the browser version — the result is a separate .txt file. Embedding an invisible text layer behind the scanned page images requires a heavier processing pipeline than is practical to run client-side. If you need a searchable PDF, desktop OCR software is the better route; the text file works well for copying and searching.
How accurate is the recognition?
On clean, straight scans of printed text at 200 DPI or more, Tesseract typically recognizes the vast majority of words correctly. Accuracy drops with low resolution, skew, stains or unusual fonts, and handwriting is generally not readable at all.
Does it work on password-protected PDFs?
No. The tool has no password prompt, so an encrypted PDF simply fails with an error message. Remove the open password first with a program that can decrypt PDFs, then run the recognition on the unlocked copy.
Can I stop a recognition that is already running?
Yes. The progress view has a Cancel button that terminates the OCR worker immediately and returns you to the start screen. Nothing is saved on cancel — partial results from pages that were already recognized are discarded along with the rest.
Want it on the go? Get the app.
All tools, fully offline, with extras the web cannot do.
