OCR — Scanned PDF to Text
Make scanned PDFs searchable
OCR for scanned documents and images. Searchable PDF, plain text, or Word output. Confidence scoring on every page. Runs entirely in your browser — nothing uploaded.
Drop scanned PDFs or images here
PDF, JPG, PNG, HEIC, WebP — or click to browse
What works: Scanned invoices, contracts, receipts, letters, forms, photos of printed text.
What doesn't: Handwriting, very old or faded documents, fax-quality scans below 150 DPI.
If your scan is borderline, drop it in anyway — we'll show you a confidence score so you can decide whether to trust the output.
What this tool does
Scanned PDFs are photographs — images of pages with no underlying text. You can't search them, select text, or copy content. OCR (Optical Character Recognition) reads each page as an image and reconstructs the text, making the document useful again.
This tool processes everything in your browser using Tesseract.js, an open-source OCR engine compiled to WebAssembly. Your files never leave your device. Each page gets a confidence score so you know how reliable the output is before you download.
What works and what doesn't
Works well
- ✓Scanned invoices and receipts
- ✓Contracts and legal documents
- ✓Letters and correspondence
- ✓Forms and applications
- ✓Phone photos of printed text (in good light)
- ✓Utility bills and bank statements
- ✓ID documents and certificates
- ✓Academic papers and reports
Doesn't work well
- ✕Handwriting (requires different AI models)
- ✕Very old or heavily faded documents
- ✕Scans below 150 DPI
- ✕Fax-quality images
- ✕Decorative or stylised fonts
- ✕Text in complex multi-column tables
- ✕Documents with heavy watermarks over text
- ✕Screenshots of curved or folded pages
How it works
Drop your document
Upload a scanned PDF or image file — JPG, PNG, HEIC, or WebP. If the PDF already has a text layer, it's extracted directly (no OCR needed).
OCR runs in your browser
Each page is processed by Tesseract, an open-source OCR engine. A confidence score tells you how reliable the output is.
Download your format
Get a searchable PDF (original layout preserved, text selectable), plain text for editing, or a Word document. Nothing is uploaded.
Which output format should I choose?
Searchable PDF
Use when: You want to keep the original document appearance
The original images are preserved exactly. An invisible text layer is added behind them — text becomes selectable and searchable in any PDF viewer without changing how the document looks.
Plain Text
Use when: You want to copy or process the text
Raw extracted text, paragraph breaks preserved. Good for pasting into other documents, running through search tools, or feeding into other software.
Word Document (.docx)
Use when: You want to edit the content
Basic paragraph structure — each page becomes a section. Opens in Microsoft Word, Google Docs, LibreOffice. No table reconstruction in this version.
Frequently asked questions
- What's the difference between OCR and just copying text from a PDF?
- A scanned PDF is just a photograph — the 'pages' are images with no underlying text data. OCR (Optical Character Recognition) reads the image and reconstructs the text. A PDF created from a word processor already has text data, so you can copy from it directly.
- How accurate is browser-based OCR?
- On clean, well-scanned documents (300 DPI or above), accuracy is typically 90–95% — good enough to trust for most uses. Phone photos of documents usually score 70–80%. Every page shows a confidence score so you know what you're working with.
- Why doesn't OCR work on handwriting?
- Tesseract, the engine used here, is trained on printed text. Handwriting recognition is a different problem that requires different models — far too large to run in a browser at useful accuracy.
- Is it safe to OCR sensitive documents like contracts or medical records?
- Yes. Everything runs entirely in your browser. Your files are never uploaded to any server — they don't leave your device. The OCR engine is downloaded once and runs locally using WebAssembly.
- What's a 'searchable PDF' and why would I want one?
- A searchable PDF looks identical to the original scanned document but has an invisible text layer underneath each page. You can use Ctrl+F to search within it, select and copy text, and screen readers can read it — all without changing the visual appearance.
- Can I OCR a phone photo of a document?
- Yes. Drop a JPG or HEIC photo directly. The tool handles rotation correction automatically. For best results, photograph in good lighting with the document flat and filling the frame.
- Why is OCR slow compared to just opening a file?
- OCR analyses each pixel of each page to reconstruct text — it's image analysis, not file reading. Expect roughly 3–4 seconds per page. A 20-page document takes around 60–90 seconds. The engine also downloads ~10 MB of language data on first use.
- Does this work offline?
- Once the OCR engine and language data have downloaded (on first use), yes — you can process documents without an internet connection.
- What languages are supported?
- English only in this version. Multi-language support (French, German, Spanish, and others) is planned for a future update.
- What if my scan is low quality?
- The confidence score on each page tells you. Below 70% usually means the output has noticeable errors. You can choose to skip low-confidence pages from the output rather than including unreliable text.
- Can I edit the text after OCR?
- Not directly in this tool — download the DOCX output and edit in Word or Google Docs. In-browser correction is planned for a future version.
- Why is the output formatted strangely?
- OCR extracts text line by line. Complex layouts — multi-column text, tables, text in boxes — get flattened. The text is there, just not always in reading order. Table extraction is a separate problem we're working on.