Question 1

Which languages are supported by this document OCR tool?

Accepted Answer

The OCR engine supports English (eng), Tamil (tam), and Hindi (hin). You can toggle between these languages depending on the primary script of your scanned document, invoice, or certificate.

Question 2

Why does scanned document OCR require an initial download?

Accepted Answer

For digital PDFs with native selectable text, extraction runs instantaneously on-device with zero downloads. For scanned documents or image files, the browser downloads the Tesseract WebAssembly engine and trained language model data (~2–4 MB) from a public CDN (jsDelivr). Once downloaded, these assets are cached locally by your browser for subsequent scans.

Question 3

What are the accuracy expectations for handwriting versus printed text?

Accepted Answer

Tesseract OCR is trained on machine-printed typography and fonts, delivering 95–99% character accuracy on clear, high-contrast printed documents (bills, receipts, certificates, printed letters). Cursive handwriting, informal notes, and heavily stylized calligraphy exhibit substantially lower accuracy and will require manual review and correction.

Question 4

Are my document contents, images, or extracted text sent to an external server?

Accepted Answer

No. While the browser downloads static language model weights from a CDN, all OCR image processing and text extraction execute entirely inside a local Web Worker on your device. Neither your original file, rendered pages, nor recognized text strings are ever uploaded to Ambiakshi or any third party.

Question 5

How can I maximize OCR extraction accuracy on scanned documents?

Accepted Answer

For best results, use scans with a resolution of at least 300 DPI, ensure the page is upright without skew, crop tightly to the text area, and avoid heavy shadows, creases, or camera glare. High contrast between crisp dark text and a clean, light background yields the highest accuracy.

Question 6

Can I export the extracted text as a structured CSV file?

Accepted Answer

Yes. In addition to copying the plain text directly to your clipboard, you can download a structured CSV file containing page numbers, line numbers, and extracted text for automated data entry, record-keeping, or spreadsheet analysis.

Extract text from a PDF

100% Client-Side Private OCR — Zero Cloud Uploads

Client-side OCR processing with CDN language models

Frequently asked

Feedback, bugs & custom requests

Related tools