OCR PDF — Get Text from Scanned Documents
Turn scanned PDFs and document photos into text in English, Thai, or both, with recognition running on your own device.
A scan is just a picture you can't search or copy from. OCR reads the letters in a scan or photo and gives you text to reuse.
How to OCR a scanned PDF (3 steps)
- Open a scanned PDF or a photo of a document.
- Choose the language: Thai + English, Thai, or English.
- Start recognition, wait a moment, and get your text.
What the tool does to improve accuracy
- An accurate model by default — you can choose a fast or an accurate model. The accurate one is a larger download and slightly slower, but reads Thai clearly better.
- Upscaling — on small pages like receipts, characters are too small to separate, so the image is enlarged according to the actual page size, not by a fixed amount.
- Sharpening — conversion to black and white with stretched contrast helps with colored backgrounds, faded receipts and unevenly lit photos.
- Receipt / table mode — left to guess a column layout, OCR often jumps around until item names mix with prices. This mode reads straight across each line for much tidier results.
- Thai clean-up — Thai stacks vowels and tone marks up to three levels, has many look-alike letters and no spaces between words, so it is harder than English. A common error with the sara am vowel is repaired automatically.
What accuracy to expect
- Printed documents scanned straight at 300 dpi or more — high; automatic mode is fine.
- Phone photos in even light — good enough to use; shoot straight and fill the frame.
- Receipts and bank slips — numbers usually right, names may not be; use receipt / invoice / table mode.
- Old, faded or faxed documents — moderate; adjust brightness and contrast at the source first.
- Handwriting — can't be read; general OCR is designed for printed text only.
No OCR is 100% accurate on every document; even paid enterprise services need a person to check. Always proofread the result, especially amounts and names.
Taking photos that OCR reads well
- Use a plain surface that contrasts with the paper, such as white paper on a dark desk.
- Shoot straight from above; tilted text distorts and gets misread.
- Avoid hand shadows and glare; a side window beats an overhead light.
- Fill the frame; larger letters mean better accuracy.
- For several pages, use the camera scanner to make one PDF, then run OCR.
Using the result
Download the text, or use the buttons below the result to convert it to Word for editing, CSV for tables in Excel (columns split by spacing), or a text-only PDF without the original scan images.
How long it takes
OCR runs entirely on your device, so it uses your own processing power. Expect about 5–10 seconds per page on a typical laptop, or 10–20 minutes for 100 pages. A progress bar shows which page it has reached. Pages are processed one at a time with memory cleared after each, so RAM use doesn't build up. On a phone, keep to around a dozen pages per run, as the device will heat up and slow down.
On first use, the tool downloads language data (size depends on language and mode, possibly several MB). Your browser keeps it, so next time recognition starts right away.
FAQ
How accurate is it?
Clearly printed documents are recognised with high accuracy. Handwriting or blurry scans may contain errors.
Is it safe for confidential documents?
Yes. OCR runs inside your own browser and your file is not sent to a server.
How long does it take?
A few seconds per page, depending on how powerful your device is.