What is the OCR tool?
OCR stands for optical character recognition: software studies a picture of a page and works out which letters are in it. That is all this tool does, except it does it inside your browser tab. Hand it a scanned PDF, a JPG or a PNG and it hands back text you can select, search and paste. No account, no install β just free OCR online, on whatever device already holds the scan.
The problem it fixes is a familiar one. A scanned document looks like text but behaves like a photograph: Ctrl+F finds nothing, you cannot lift a quote out of it, and a screen reader has nothing to announce. Retyping a twelve-page article by hand is an hour you never get back. Recognition turns that flat image back into characters β either as a plain .txt file, or as a searchable PDF that looks identical to the original but has an invisible text layer sitting on top of each page.
The recognition engine is Tesseract compiled to WebAssembly, running in your tab. The page image is never posted anywhere, which matters when the scan is a payslip, a medical letter or an ID page. Two honest trade-offs come with that: the language pack (a few megabytes) is fetched once from a CDN on your first run, and your own processor does the reading, so long documents take real time β think a few seconds per page on a laptop and noticeably slower on a phone.
How to OCR a scanned PDF or image
- Drop a scanned PDF, JPG or PNG onto the upload area, or click it to browse. One file per run.
- Set Language to match the document β English, Bengali, Hindi, Arabic, Spanish, French, German, or the combined English + Bengali option for mixed pages.
- Choose Output: plain text if you want to edit or quote it, or Searchable PDF to keep the original appearance and gain a text layer.
- For a PDF, fill in Pages to read (leave it empty for all, or type something like 1-3, 5) and pick a Detail level β 200 dpi suits most scans, 300 dpi helps with small print.
- Watch the preview while you adjust Clean-up before reading: greyscale with a contrast boost is the default, and black & white with an adjustable cut-off rescues faint or grey pages.
- Press Run OCR, read the recognised text in the box that appears, then choose Download result.
Key features
- Two kinds of output β a plain text file, or a searchable PDF built from the page image with the recognised words drawn invisibly in place.
- Eight language settings β including a dual English + Bengali mode for notes and forms that switch script mid-line.
- Confidence highlighting β after a run every recognised word is boxed on the preview in green, amber or red, and the status line reports the average confidence for the whole job.
- A preview that tells the truth β the same clean-up code draws the preview and prepares the image sent to the recogniser, so what you see is literally what gets read.
- Manual clean-up controls β contrast from 80% to 220%, and a black-and-white cut-off from 20% to 85% for scans that are too pale or too dark.
- Page ranges and resolution β read only the pages you need at 150, 200 or 300 dpi instead of grinding through a whole report.
- Text you can check first β the result appears in a box on the page, so you can see how it went before you save anything.
Benefits of using this OCR tool
- Sensitive scans stay put. Contracts, bank letters and ID pages are recognised on your device, so there is no upload to trust and nothing to delete afterwards.
- No page caps or daily quotas. Most free OCR sites stop you after a handful of pages; here the only ceiling is how long you are willing to wait.
- It keeps working offline. Once the language pack has been cached, you can drop the connection and carry on recognising pages on a train or a plane.
- Clean output, no branding. The text file is just text and the searchable PDF carries no watermark, footer or trial stamp.
- You can see where it struggled. Red and amber boxes point straight at the words worth double-checking, instead of leaving you to spot silent errors later.
Who should use this tool?
Students digitising notes
Photograph the handout or the library page, run it through at 300 dpi, and you have text you can drop into your own notes, reorganise, or feed to a summariser before an exam. Plain-text output is the right pick here β it pastes cleanly into any editor.
Researchers with scanned papers
Archive PDFs and interlibrary scans are usually image-only, which makes a literature review painful. Convert them to searchable PDFs and your reference manager, your reader's search box and your quote-hunting all start working. Use page ranges to read just the methods section if that is all you need.
Offices handling receipts and invoices
A photo of a receipt is not data. Recognise the text, pull out totals and reference numbers, and paste them into the spreadsheet without typing. Because nothing leaves the building, this also passes the kind of policy that forbids uploading supplier documents to third-party converters.
Anyone building a searchable archive
Old letters, certificates, minutes, manuals β anything you scanned years ago and can no longer find. Running each file through as a searchable PDF keeps the original look while making the whole folder findable, and it makes the documents readable by assistive technology too.
Frequently asked questions
Is this OCR tool free, and are there hidden limits?
It is free with no signup and no page counter. All the computing happens on your machine, so there is no server bill behind it and nothing to upsell you.
How accurate is the text recognition?
Accuracy depends almost entirely on the scan. A flat, well-lit 300 dpi scan of printed text is usually near-perfect; a crooked phone photo with shadows, or a fax-quality page, will produce mistakes. Raising Detail, switching clean-up to black & white and nudging the cut-off often fixes a bad first result.
Do my scanned files get uploaded to a server?
No. Pages are rendered and recognised inside the browser tab. The only network request is the one-off download of the language model from a CDN β your document is not part of it.
Which languages can it extract text from?
English, Bengali, Hindi, Arabic, Spanish, French and German, plus a combined English + Bengali mode. Pick the closest match to your page: choosing the wrong language is one of the most common causes of garbled output.
What exactly is a searchable PDF here?
Each page image is placed in a new PDF with the recognised words drawn on top at zero opacity, so the document looks unchanged but text can be searched and copied. Note that the hidden layer uses a standard Latin font, so for Bengali, Hindi or Arabic pages the plain-text output is the reliable choice.
Why does a long document take so long?
Because recognition runs on your own CPU rather than a server farm. Budget a few seconds per page on a desktop and more on a phone. If you only need part of a file, set a page range instead of reading all of it, or drop Detail to 150 dpi for clean, large print.
Can it read handwriting?
Not reliably. The engine is trained on printed type, so neat block capitals sometimes come through while ordinary cursive usually does not. Treat handwriting as out of scope rather than something to fight with the sliders.
Can I run OCR on a password-protected PDF?
Not while it is encrypted β the file cannot even be opened for rendering. Remove the password with the Unlock PDF tool first, then bring the unlocked copy back here.
Related tools you might need
Got a scan you cannot search? Scroll up, drop it in and run free OCR online β your file stays on your device from start to finish.
Run OCR now