How to extract text from a scanned PDF (OCR)

YYou receive a scanned contract and try to highlight a clause to paste into an email — nothing happens. The cursor passes right over the text without reacting. That's because the "text" in the PDF is actually a photograph of a piece of paper. To any program, you're just selecting a JPG.

OCR (Optical Character Recognition) is the technology that lets a machine "read" those images and convert them into real text. The result: a document you can search with Cmd/Ctrl-F, copy passages from, edit in Word, index for search, or feed to an LLM.

3-second test: open the PDF and try to select a word with your mouse. If it highlights cleanly word by word, the PDF has real text and doesn't need OCR. If you can only draw a rectangle on top of it, it's an image — you need OCR.

Another tell: hit Cmd/Ctrl-F and search for a word you KNOW is in the document. If it doesn't find it, the content is an image.

• —Old digitized contracts — to extract clauses, dates, amounts
• —Scanned receipts — to populate expense spreadsheets
• —Books and academic papers — to cite passages, translate
• —HR documents — IDs, payslips, certificates for record-keeping
• —Medical history — to digitize old patient records
• —Field research — partially handwritten survey forms

1. Upload the scanned file

Works with PDF, JPG, PNG and TIFF. You can upload multi-page documents — OCR processes everything in one shot and preserves the order.

2. Pick the content language

We support English, Portuguese, and Spanish. OCR uses different models per language — picking the wrong one tanks accuracy. If your document is multilingual (a bilingual report, for instance), process each part separately.

3. Choose output format

• —Searchable PDF — keeps the original visual but adds an invisible text layer on top. You can Cmd/Ctrl-F, copy and paste normally.
• —Plain text (.txt) — just the extracted content, no formatting. Great for spreadsheets, importing into systems, feeding to AI.
• —Word (.docx) — converts with basic formatting preserved (paragraphs, alignment). Good for editing.

4. Process and download

OCR is slower than other conversions (each page takes 2-10 seconds depending on resolution). When it finishes, you download the file in your chosen format.

OCR accuracy depends heavily on the source image quality. Some tips:

• —Scan at 300 DPI minimum — below that, small letters blur
• —Make sure the page is straight — tilted scans and angled shots confuse the recognition
• —Clean stains and folds before scanning — marks become random characters
• —Prefer white background and black ink — high contrast = better reading
• —Avoid screen captures — moiré and pixelization hurt accuracy

OCR unlocks several follow-ups:

• —OCR + Compress — after OCR the PDF gets MUCH lighter (text weighs far less than image)
• —OCR + Word — export to .docx for editing and review
• —OCR + Excel — if the document is tabular, OCR + Excel converter splits into columns
• —OCR + ChatGPT — drop the extracted text into AI to summarize, translate, or analyze