A scanned PDF is a stack of photographs of paper. You can read it, but your computer cannot: search finds nothing, you cannot copy a sentence, and a screen reader has nothing to read aloud. OCR (optical character recognition) fixes that by finding the letters in each picture and writing them into the PDF as real text.
This tool runs OCR on our server with Tesseract, the widely used open-source OCR engine, and hands you back a PDF that looks exactly like your scan, with an invisible text layer placed over the words. Here is how to use it, how to scan so OCR gets the words right, and what to do with the text afterwards.
How to OCR a PDF
- Drag your scanned PDF onto the box, or click the box and choose it. A JPG, PNG or TIFF image of a page works too; it becomes a one-page PDF.
- Look at the note under the file name. For a PDF it shows the page count, and it warns you if the file seems to contain text already.
- Choose the language of the text. English is selected by default. Tap more languages if the document mixes them, up to three at once.
- Click Run OCR. The progress bar shows the upload, your place in line and the recognition itself. Longer documents take longer, because every page is read separately.
- The searchable PDF downloads on its own when it is ready. If your browser blocks it, click Download PDF.
- Open it and press Ctrl+F (Cmd+F on a Mac). Search for a word you can see on a page. If it is found and highlighted, OCR worked.
What OCR changes in your PDF, and what it does not
People often expect OCR to turn a scan into an editable document. That is not what a searchable PDF is, and it helps to know the difference before you start.
| After OCR you can | After OCR you still cannot |
|---|---|
| Search the document with Ctrl+F or Cmd+F | Edit the words directly on the page |
| Select and copy text into an email or document | Change the font or layout |
| Find the file by its contents with Spotlight on a Mac, or in cloud drives and document systems that index PDF text | Fix a typo in the original paper |
| Have a screen reader read the page aloud | Get perfectly accurate text from a blurry or handwritten page |
The picture of each page is kept as it is. The text layer sits underneath what you see, lined up with the words, which is why selecting text highlights the right spot on the scan. Nothing is redrawn, so signatures, stamps and handwritten notes stay exactly as scanned.
If you need to extract all of the text at once, run the searchable PDF through PDF to Text, which reads the new text layer in your browser.
Choosing the right language
OCR does not just see shapes; it matches them against what words in a language look like. Telling it the right language makes a real difference, especially for accented letters (é, ü, ñ, ç), different alphabets and right-to-left scripts.
- One language: select only that one. It is the fastest and most accurate setting.
- Mixed documents: an English invoice with a French address, or a Spanish contract with English product names. Select both.
- Up to three: each extra language makes OCR a little slower and gives it more ways to misread a word, so pick only what is really on the page.
The list covers more than 30 languages, including Arabic, Persian, Urdu, Hindi, Bengali, Chinese (simplified and traditional), Japanese, Korean, Thai and Vietnamese, along with the main European languages.
Scan so OCR gets the words right
The quality of the scan matters more than anything else. Most OCR errors come from pages that are too small, too dark or crooked. These settings give the best results:
| Setting | Use | Why |
|---|---|---|
| Resolution | 300 DPI | Small print below 200 DPI loses the details that separate letters like e and c |
| Color mode | Grayscale, or black and white for clean text pages | Clear contrast between ink and paper; color only adds file size |
| Page position | Straight and flat on the glass | Curved lines near a book's spine read badly |
| Phone scans | A scanner app (Notes on iPhone, Google Drive on Android) instead of a plain photo | Scanner apps flatten, crop and boost contrast automatically |
The tool does not rotate pages. A page that is sideways or upside down will give poor results, so turn it the right way first with Rotate PDF. Slightly crooked scans are usually fine.
Scanning with a phone
You do not need a scanner. Both major phone systems have a document scanner built in that produces a clean PDF:
- iPhone: in the Files app, tap the three dots at the top and choose Scan Documents. The same scanner is in Notes. Hold the phone straight over the page; it captures automatically when the page is steady, and you can add more pages before you tap Save.
- Android: in the Google Drive app, tap the + button and choose Scan. Crop each page if the edges are off, then save it as a PDF.
Scan in good daylight or under an even lamp, and keep shadows (including your own hand and phone) off the page. Then upload the PDF here for OCR.
What OCR reads well, and where it struggles
- Typed and printed text: letters, invoices, contracts, books, forms filled in by computer. Expect very good results from a clean 300 DPI scan.
- Tables: the text in every cell is recognized and becomes searchable. The table structure stays a picture; to get rows and columns into a spreadsheet, try PDF to Excel on the searchable copy and check the result.
- Small print and footnotes: fine at 300 DPI, unreliable at lower resolutions.
- Stylized fonts, logos and text on photos: often missed or misread. They are rarely the part you need to search.
- Handwriting: not reliable. Tesseract is built for printed text. Neat block capitals sometimes work; joined-up handwriting usually comes out as nonsense or nothing.
- Faint photocopies, fax pages and stamps over text: expect gaps. Rescan from the original if you have it.
Warning: OCR text can contain errors that are invisible when you look at the page, because you see the scan, not the text layer. Before you rely on copied numbers, such as an account number or a total, compare them with the scan.
If the PDF already has text
Many PDFs are not scans at all. Anything exported from Word, Google Docs, an invoicing app or a website already has real text. The quick test: open the PDF and try to select a sentence with your mouse. If the words highlight one by one, the page has text and does not need OCR.
Some PDFs are mixed: a typed report with a few scanned pages added at the end, for example. The tool notices when a PDF contains fonts and warns you. Pages that already have text are left as they are, and only the pages that are pictures get OCR, so running a mixed file through it is safe.
Troubleshooting
- "No text was found": the pages are blank, photos without text, or handwriting. It also happens when the wrong language is selected for a non-Latin script, such as Arabic or Chinese with only English selected.
- Search finds some words but not others: the missed words were misread. Search for a shorter part of the word, or rescan the page at 300 DPI.
- Copied text has a line break after every line: that is how the lines sit on the paper. Paste into a plain text editor and join the lines, or paste into Word and use Find and Replace on paragraph marks.
- Columns come out mixed together: in magazine-style layouts, select one column at a time when copying. Searching is not affected.
- Accented letters are wrong: add the document's language to the selection. English alone does not expect é, ø or ğ.
- The file did not upload: check the size limit in the box. Split a very large scan with Split PDF and OCR the parts.
Example: a 40 page scanned lease
A landlord emails a 40 page lease as one scanned PDF of 22 MB. You want to find every mention of "deposit" and "notice" before you sign.
- Open the PDF and try to select a word. Nothing highlights, so it is a scan and needs OCR.
- Drop it into the box. The note shows 40 pages made of images.
- The lease is in English, so keep English selected and click Run OCR.
- When the searchable copy downloads, open it and search for "deposit". Each hit is highlighted on the scan, page by page.
- Copy the relevant clauses into your notes and compare any amounts with the scan before you rely on them.
The whole job takes minutes instead of reading 40 pages twice, and the searchable copy stays useful for the length of the lease.
Browser OCR or server OCR?
The site has two OCR tools, and they suit different jobs:
| Image to Text | OCR PDF (this page) | |
|---|---|---|
| Where it runs | In your browser, nothing uploaded | On our server |
| Input | Images | Scanned PDFs of any length, or images |
| Output | Plain text to copy | A PDF that looks like the scan, with searchable text |
| Best for | Grabbing text from one screenshot or photo | Archiving, sharing and searching whole documents |
What happens to your file
OCR needs more computing power than a phone or browser comfortably has, so the file goes to our server. It is sent over HTTPS with only the tool name and your language choice, processed, and deleted: the upload as soon as the job ends, the searchable PDF 30 minutes after it is ready. After the automatic download, the Download button keeps a copy in your tab so you can save it again until you leave the page.
Common uses
Paper archives and receipts
Scan once, OCR once, and every receipt, warranty and contract becomes findable by a word you remember from it, such as the store name or an amount. Name the files with a date and a short description too; search works best when both are in place. On a Mac, Spotlight searches inside searchable PDFs, so a plain folder of scans becomes a document archive without extra software.
Supplier catalogs and price lists
Suppliers still send scanned catalogs and price sheets. After OCR you can search for a product code instead of paging through 80 scans, and copy names and prices into your own sheet. If you compare those products with what sells online, AM Jarvis Marketplace Research checks prices and demand across Amazon, eBay, Etsy and other marketplaces.
Accessibility
A scanned PDF is silent for people who use screen readers. A searchable copy gives their software real text to read, which is the least you can do before posting a scanned document on a website or sending it to a large group.