OCR Text Recognition
Run optical character recognition on scanned PDFs completely client-side. Sensitive documents never leave your machine.
PDF Vault runs OCR on scanned PDFs and images entirely in your browser using a local Tesseract engine, so sensitive scans are never uploaded to a cloud OCR service.
OCR is the step that turns a picture of words into words. Without it, a scanned contract is an image: you cannot search it, copy from it, or feed it into anything else.
Conventionally this is a server job, which is precisely the problem, because the documents that most need OCR — medical records, court exhibits, historical case files, tax paperwork — are the ones least appropriate to upload to a third-party processor.
PDF Vault runs a Tesseract-based recogniser compiled to WebAssembly, locally. It supports over 40 languages and produces text you can copy, export, or keep as a searchable layer. The first run downloads the language model; after that it works offline.
How to OCR a scanned PDF offline
- Load the scanned PDF or image.
- Pick the document's language so the recogniser uses the right model.
- Run the recognition pass and wait — this is the most computation-heavy tool here, and it is using your CPU.
- Copy the extracted text, export it, or keep the searchable output.
What you get
- 40+ languages. Tesseract language models covering most Latin, Cyrillic, and CJK scripts.
- Fully local. No cloud OCR API ever receives your scan.
- Searchable output. Turn an image-only document into one you can search and copy from.
- Offline after first run. Once the language model is cached, recognition works with no connection.
Frequently asked questions
How accurate is browser-based OCR?
On a clean 300 DPI scan of printed text, accuracy is high and comparable to server engines. Accuracy drops on low-resolution scans, skewed pages, unusual fonts, and handwriting — the same factors that affect any OCR engine.
Why is OCR slower than the other tools?
Because it is genuinely heavy computation and it is running on your CPU rather than a datacentre GPU. That is the tradeoff for the document never leaving your machine.
Does it read handwriting?
Tesseract is built for printed text. Handwriting recognition is unreliable and should not be depended on.
Which languages are supported?
Over 40, selectable before the run. Choosing the correct language materially improves accuracy.