PDF OCR: Make Scanned PDFs Searchable
Add a searchable text layer to scanned PDFs and images. Runs in your browser.
About this pdf ocr: make scanned pdfs searchable
This PDF OCR tool turns scanned PDFs and photos of pages into searchable PDFs you can select, copy and search with Ctrl+F. It reads English and Bengali (Bangla), and it runs entirely in your browser.
A scan is just a picture of text, so your computer can't search it. OCR (optical character recognition) finds the letters in that picture. The tool then lays the recognized words over each page as an invisible text layer, exactly where they appear. The page looks the same as before. It simply becomes searchable.
You also get the plain text as a TXT file and an hOCR file with the position of every word. Since nothing is uploaded, it's a sensible choice for ID cards, contracts and medical letters.
Features
Searchable PDF output
The original page is kept untouched, vectors and all, and a hidden text layer is added on top. Search, select and copy work in any PDF reader.
English, Bengali or both
Pick the language before you start. Both is for documents that mix Bangla and English, such as forms and official letters.
PDFs or images
Open a scanned PDF, or several JPG, PNG or WebP photos of pages. Photos become one multi-page searchable PDF in file order.
Skips pages that already have text
Mixed PDFs often have some typed pages and some scanned ones. By default only pages without text are read, which saves time. Untick the option to OCR everything.
TXT, hOCR and copy
Download the text with a header for each page, or hOCR (HTML with word boxes and confidence scores) for further processing. A per-page confidence score flags pages worth checking.
Page progress and cancel
Each page is drawn at 300 dpi and read in a background worker, so the page stays responsive. You can stop at any time and keep the pages already done.
How to use the pdf ocr: make scanned pdfs searchable
- 1Choose a scanned PDF or photos of pages. A sample scan is loaded to try.
- 2Pick English, Bengali or both, and the pages to read.
- 3Click Make searchable, then download the PDF, TXT or hOCR.
When to use it
Find things in old scans
Make a box of scanned invoices or letters searchable, so a client name or amount turns up with Ctrl+F.
Copy text out of a scan
Pull a paragraph from a scanned book chapter or circular without retyping it. Paste it into a document, or into the word counter if you have a length limit.
Bangla documents
Government notices, land papers and school certificates in Bengali can be read and searched, which most OCR services handle poorly or not at all.
Phone photos of paper
Photograph each page, open all the photos here, and get one searchable PDF. Or turn the photos into a PDF first if you want to reorder or crop pages.
What is OCR in a PDF?
OCR software looks at an image, finds lines of text, and predicts the characters. This tool uses Tesseract, the open-source engine originally developed at HP and later maintained by Google, compiled to WebAssembly. Its LSTM neural network reads whole lines at once, which copes better with joined Bengali letters than older character-by-character methods.
A searchable PDF keeps the scan as the visible layer and adds the text with PDF's invisible text rendering mode. That's why the page doesn't change when you view or print it, yet a reader can find and highlight the words.
How accurate is it, and how long does it take?
On clean pages the results are very good. In our tests, a typed English A4 page at 300 dpi was read with every one of its 82 words correct. A Bengali page scored 36 of 37 words, with one stray vowel sign added to a word.
Real scans are messier. Skewed pages, shadows, low resolution, stamps and decorative fonts all cost accuracy. Handwriting isn't supported; the models are trained on printed text.
Speed depends on your device. In our tests a full A4 page at 300 dpi took about 9 to 11 seconds in Chrome on a desktop PC. Phones are slower. A 100-page book can take 15 minutes or more, so the tool warns you above 50 pages.
How to get better OCR results
Most problems start at the scanner. These habits help the most:
- Scan at 300 dpi. Tesseract's own documentation says it works best at 300 dpi or more.
- Keep the page straight and flat. A few degrees of tilt is fine; a curled book spine is not.
- Use even light without shadows. Black text on white paper is ideal.
- Pick the right language. English mode can't read Bengali script, and vice versa.
What gets downloaded the first time?
OCR needs an engine and language models. They come from this website, not a third party: about 7 MB for English, 5.5 MB for Bengali, or 8.5 MB for both. Your browser caches them, so the next run starts in a second or two.
Why use Mavros
| Mavros | Typical upload sites | |
|---|---|---|
| Where the file goes | Stays in your browser | Uploaded to a server |
| Bengali (Bangla) | Yes, alone or mixed with English | Often missing |
| Page limits | None, with a time warning over 50 pages | Often 3 to 10 free pages |
| Watermark | Never | Sometimes on free output |
| Outputs | Searchable PDF, TXT and hOCR | One format, others paid |
Tips
- Check pages flagged with low confidence. They usually have faint print or a stamp across the text.
- Searchable PDFs are a little larger than the scan. Compress the searchable PDF afterwards if size matters; the text layer survives.
- For a long document, try a page range like 1-5 first to check quality and timing.
- A password-protected scan has to be opened first. Remove the password with the password you know, then run OCR on the copy.
Questions
How do I make a scanned PDF searchable?
Open the PDF here, pick its language and click Make searchable. The tool reads each page and adds an invisible text layer. Download the new PDF and search it with Ctrl+F in any reader.
Does OCR change how my PDF looks?
No. Every page keeps its original appearance. The recognized text is invisible and sits exactly over the printed words, so only search, selection and copy change.
Is my document uploaded?
No. The OCR engine runs in your browser. Only the engine and language files are downloaded from this site; your document never leaves your device.
Can it read Bengali (Bangla) text?
Yes. Choose Bengali, or Both for mixed documents. Clean print reads well; very old prints, unusual fonts and broken conjunct letters cause more errors than English does.
Can it read handwriting?
Not reliably. The models are trained on printed text. Neat block capitals may partly work, but cursive handwriting will mostly come out wrong.
How do I know if a PDF already has text?
Try to select a word, or press Ctrl+F and search for one. If nothing highlights, the page is an image. This tool also checks each page and skips ones that already have text.
What is hOCR?
hOCR is an open format that stores OCR results as HTML, with each word's position on the page and a confidence score. Developers use it to build search indexes or to correct OCR output.
Why is the OCR text slightly wrong?
Low resolution, tilt, shadows or fancy fonts confuse the engine. Rescan at 300 dpi with the page flat if you can, and check the pages flagged as low confidence.