The developer released react-native-glyph-ocr, a library for on-device text recognition in React Native. It utilizes Apple Vision for iOS and PaddleOCR for Android, addressing the lack of Cyrillic support in Google ML Kit.
On Android, the library uses the PaddleOCR PP-OCRv5 mobile model with the eslav package. Performance tests on food labels showed that PaddleOCR recognizes 77–83% of ingredients, while Apple Vision on iOS achieves 82–97%. In comparison, ML Kit reaches only 13–28%.
The library supports custom model management via Hugging Face manifests or embedded assets. It includes a scanner hook that evaluates image quality based on motion, blur, and sharpness before triggering recognition, ensuring efficient processing on mobile hardware.
The project requires React Native 0.85 or newer, iOS 16+, and Android 7+ (with Android 10 required for camera frame buffers). The current version is 0.3.1, and the developer plans to further optimize performance and expand testing beyond food labels.
