Detect and recognize
PP-OCRv6 ONNX locates text while a mixed-script recognition model reads the line content.
Private browser tool
Turn Khmer scans into reviewable text.
Recognize PDF and image documents, inspect uncertain lines, and export reconstructed searchable PDFs. Inference and document processing stay in your browser.
Recognition with review
Keep the source image and the recognized text connected.
Diagnose and correct
Viterbi segmentation, spelling diagnostics, and ranked lexicon suggestions focus review on uncertain or invalid lines.
Reconstruct documents
Export page images with an invisible Unicode text layer for search, selection, and copy in compatible PDF viewers.
Local workflow
Your source files are never uploaded.
The application loads models and runs detection, recognition, language analysis, review, and document export on the device.
- retain immutable raw recognition during review
- keep page images in reconstructed output
- choose whether text is copied or saved as a file
PDF or imagelocal file
↓
Detect and recognizeONNX in browser
↓
Segment and reviewUnicode text
↓
Exporttext or searchable PDF
The exported PDF preserves page images and adds searchable Unicode, but it is not fully editable. Firefox's built-in PDF viewer is incompatible with the custom semantic text layer. Use Chromium or a compatible desktop PDF reader. PDF/A conformance is not claimed.
No installation required
Recognize a Khmer document privately.
Open the tool in a modern browser. WebGPU is preferred when available, with a WASM fallback.