Live Demo

How It Works

1. Client-Side Processing: The OCR inference pipeline executes directly on your machine using WebAssembly.
2. Interactive Bounding Boxes: Hover over any detected bounding box on the image canvas to inspect the recognized text and confidence percentage. Click any box to copy its text.
3. Flexible Granularity: Switch between line-level grouping (lines) and individual word tokens (words), with 2D spatial layout reconstruction available in the Spatial Layout sub-tab.
4. Data Export: Export results instantly to Clipboard, CSV, or structured JSON.

|

Model:

PP-OCRv6 Tiny (~6 MB) PP-OCRv6 Tiny (INT8) (~3.4 MB) PP-OCRv6 Small (~30 MB) PP-OCRv6 Small (INT8) (~16 MB) PP-OCRv6 Medium (~138 MB) PP-OCRv6 Medium (INT8) (~76 MB) PP-OCRv5 Mobile (~12 MB) PP-OCRv5 Server (~166 MB) PP-OCRv4 Mobile (~16 MB) PP-OCRv4 Server (~196 MB)

Ultra-lightweight

Settings

Run OCR

Sample Image: InvoicePassportHandwritten+ Upload File

Tip: Hover over bounding boxes to highlight recognized text. Click to copy.

lineswords

Copy All

Items (0) Spatial Layout CSV JSON Config

No text detected or score below threshold.

No run yet

Details

Integration Guides

• Use in Web Apps (vanilla JS / TS)
• Use in React Applications