An Offline OCR Tool That Recommends Its Own Replacement

What it does This is a Windows desktop fork of the Tianruo OCR tool that extracts text from screen captures using local Chinese-lite and PaddleOCR ONNX models instead of cloud APIs. It also pipes results through several translation services, including an optional self-hosted Python backend for offline translation. The author—a mechanical engineering student—describes the codebase as largely assembled from existing libraries and forum searches.

The interesting bit The README doubles as a confession: the maintainer admits the project is mostly glue, recommends the better-maintained Umi-OCR, and yet still compiled a custom ONNX Runtime DLL to keep Win7 compatibility alive. It is a rare case where a developer’s self-awareness outpaces the software’s polish.

Key highlights

• Fully offline text recognition using local PaddleOCR v3 and Chinese-lite ONNX models
• Custom-compiled ONNX Runtime to support Windows 7 through Windows 10 (64-bit)
• Hot-swappable recognition backends without restarting the application
• Optional offline translation via a self-hosted Python Flask server (requires ~1 GB of model downloads)
• Barcode support added in v1.3.9 through ZXING

Caveats

• Explicitly unmaintained; the author recommends migrating to Umi-OCR
• Offline translation setup demands manual Python environment configuration and significant RAM
• The author cautions that their non-programmer background means many issues are effectively “won’t fix”

Verdict Worth a spin if you need local OCR on older Windows machines and can tolerate rough edges. For most users, the author is right—Umi-OCR is probably the smoother choice.