This Khmer OCR tool reads the Khmer writing in a photo, screenshot or scanned document and turns it into real Khmer Unicode text that you can copy, edit, search and paste into Word, Telegram or Facebook. It uses Tesseract 5, an open-source OCR engine with a neural network trained for Khmer, together with English so that names, numbers, prices and web addresses in a Khmer text come out right too. Several images or a whole PDF can be read in one go.
Before reading, each picture is cleaned up automatically: a tilted photo is straightened, small text is enlarged, shadows and uneven light are evened out and grain is removed. On our Khmer test set this lowered the character error rate from about 9.5 % to 2 %. Afterwards the text gets a Khmer clean-up: vowel signs and subscripts are put in the standard Unicode order and stray spaces are removed, and one click opens the result in the Khmer spell checker. By default everything runs in your browser and nothing is uploaded.
How to use the Khmer OCR tool
- Click Choose images or PDFs, drop files onto the box, paste a screenshot with Ctrl+V, or tap Take a photo on a phone.
- Leave the language on Khmer with Also read English ticked (best for most Khmer documents), and the layout on Block of text.
- Click Extract text. The first time, the engine and the Khmer data (about 1 MB) are downloaded and kept for next time.
- Check the result: the Confidence view marks words the engine is unsure about in yellow and red. Correct them in the text box.
- Click Copy text or Download TXT, or Check Khmer spelling to review the text in the spell checker.
Why use ToolOrca Khmer OCR
- Khmer Unicode output with English, numbers and URLs
- Automatic straightening, enlarging and lighting fix
- Fixes vowel and subscript order, removes stray spaces
- Photos, screenshots and scanned or digital PDFs
- Several files at once, one combined text or a ZIP
- Confidence view that marks words to check
- Runs in your browser; server option for many pages
- Free, unlimited, no sign-up and no watermark
How accurate is Khmer OCR?
On clear printed text, such as a document from Word, a book page or a sharp screenshot, Tesseract reads Khmer very well: on our test set of Khmer paragraphs in Noto Sans Khmer, Battambang, Khmer OS, Hanuman and Siemreap fonts, about 99 of 100 characters were right on clean images and on scans. Blurry, small, tilted phone photos are harder; the automatic clean-up brought those from a mostly unreadable result to about 85 % of characters right. Handwriting and decorative fonts such as Moul are not supported well.
Typical mistakes are look-alike letters (ព and ឰ, ប and ឋ), a missing small vowel sign under a letter such as ុ or ូ, and a wrong subscript. They are usually marked yellow or red in the confidence view, so you can find them quickly. We also tried an open-source Khmer recognition model (Kiri OCR) and found Tesseract more accurate on our test pages, so Tesseract is what this tool uses.
Tips for better results
Light and focus matter most. Photograph the page in daylight or under a bright lamp, without your shadow on it, and hold the phone flat above the page so the lines are straight. Tap the screen to focus before taking the photo. A screenshot is always better than a photo of a screen.
Fill the picture with the text. Crop away the table, hands and other pages, or take the photo closer. If the text is in columns, choose Automatic (columns, mixed blocks) as the layout; for one line such as a sign or a label, choose Single line. If a result looks wrong, try Black and white clean-up for faded print, or Off for clean digital images.
PDF files: a PDF made from Word already contains its text, and the tool takes it directly, which is faster and exact. Old Khmer PDFs typed with Limon or ABC fonts come out as Latin letters; read those as images instead, or convert them with the Limon to Unicode tool.
Khmer clean-up after OCR
Khmer letters are built from a base consonant with subscripts, vowel signs and other marks, and these must be stored in a fixed order to display and search correctly. The clean-up puts every cluster into the standard order (for example a subscript before a vowel sign), removes doubled marks and zero-width characters inside clusters, deletes the space Tesseract sometimes leaves before ។ and drops stray symbols such as ! or | read inside a Khmer word.
We tested an automatic spelling correction pass as well, but on our test set it changed more right words than wrong ones, because names and rare words are not in the dictionary. That is why spelling is a separate step: Check Khmer spelling opens the text in the Khmer spell checker, where you choose each correction yourself.
Frequently asked questions
Is this Khmer OCR free?
Yes. It is free with no sign-up, no page limit when it runs on your device and no watermark. The server option has a fair-use limit per minute and per day.
Are my images uploaded to a server?
Not by default. With On this device (the default) the OCR engine runs in your browser and your files never leave your phone or computer; only the engine and the language data are downloaded once from a public CDN. If you choose On our server, each page is cleaned up in your browser and the prepared page image is uploaded over an encrypted connection, read by the same engine and deleted right after the text is sent back.
Can it read Khmer and English in the same image?
Yes. Keep Also read English words and numbers ticked; on our mixed Khmer–English test images this cut the errors by more than half compared with Khmer alone. Untick it only for text that is purely Khmer.
Can it read handwritten Khmer?
Not reliably. The engine is trained on printed fonts, so neat handwriting gives a rough result at best. For printed text, typed documents, screenshots, signs and book pages it works well.
Why are some words highlighted?
The engine gives every word a confidence score. In the Confidence view, yellow words (60–80 %) are worth a check and red words (below 60 %) are probably wrong. Hover over or tap a word to see its score.