Expo Text Extractor

Fork Notice: This is a fork of pchalupa/expo-text-extractor with added multi-language support for Chinese, Japanese, Korean, and Devanagari scripts.

Expo Text Extractor is a library that enables text recognition (OCR) using Google ML Kit on Android and Apple Vision on iOS.

Platform Compatibility

| Android Device | Android Emulator | iOS Device | iOS Simulator | Web | | -------------- | ---------------- | ---------- | ------------- | --- | | ✅ | ✅ | ✅ | ✅ | ❌ |

Supported Languages

The library supports multiple scripts and languages:

| Script | Languages | Android | iOS | | ---------- | --------------------------------------------------- | ------- | --- | | Latin | English, Spanish, French, German, Italian, etc. | ✅ | ✅ | | Chinese | Simplified Chinese, Traditional Chinese | ✅ | ✅ | | Japanese | Japanese (Hiragana, Katakana, Kanji) | ✅ | ✅ | | Korean | Korean (Hangul) | ✅ | ✅ | | Devanagari | Hindi, Sanskrit, Marathi, Nepali, etc. | ✅ | ✅ |

Demo

Installation

To get started, install the library:

npx expo install @zhanziyang/expo-text-extractor

Or with npm/yarn:

npm install @zhanziyang/expo-text-extractor
# or
yarn add @zhanziyang/expo-text-extractor

Ensure your project is running Expo SDK 52+.

API Documentation

Check the example app for more details.

isSupported

A boolean value indicating whether the current device supports text extraction.

import { isSupported } from '@zhanziyang/expo-text-extractor';

if (isSupported) {
  console.log('Text extraction is supported on this device.');
}

extractTextFromImage(uri, options?)

Extracts text from an image and returns the recognized text as an array.

import { extractTextFromImage, TextRecognitionScript } from '@zhanziyang/expo-text-extractor';

// Basic usage (Latin script by default)
const texts = await extractTextFromImage('file:///path/to/image.jpg');

// With a specific script (Android)
const chineseTexts = await extractTextFromImage('file:///path/to/image.jpg', {
  script: TextRecognitionScript.CHINESE,
});

// With specific languages (iOS)
const japaneseTexts = await extractTextFromImage('file:///path/to/image.jpg', {
  languages: ['ja-JP', 'en-US'],
});

Parameters

| Parameter | Type | Description | | --------- | -------------------- | ---------------------------------------------- | | uri | string | The URI of the image to extract text from. | | options | RecognitionOptions | Optional. Recognition options (see below). |

Recognition Options

| Property | Type | Platform | Default | Description | | ------------------------------ | ------------------------- | -------- | ----------- | --------------------------------------------------------------------------------------------- | | script | TextRecognitionScript | Android | LATIN | The script type to use for recognition. | | languages | string[] | iOS | - | Array of BCP-47 language codes (e.g., 'en-US', 'zh-Hans', 'ja-JP'). | | automaticallyDetectsLanguage | boolean | iOS 16+ | false | Automatically detect language without specifying languages. | | usesLanguageCorrection | boolean | iOS | true | Apply language correction to improve recognition accuracy. | | customWords | string[] | iOS | - | Custom words (proper nouns, technical terms) to help with recognition. | | minimumTextHeight | number | iOS | ~0.03 | Minimum text height as fraction of image height (0-1). Lower values detect smaller text. | | recognitionLevel | RecognitionLevel | iOS | ACCURATE | Recognition level: ACCURATE (slower, better) or FAST (faster, less accurate). |

TextRecognitionScript Enum (Android)

| Value | Description | | ------------- | ------------------------------------------------------------------ | | LATIN | Latin script - English, Spanish, French, German, Italian, etc. | | CHINESE | Chinese script - Simplified and Traditional Chinese | | DEVANAGARI | Devanagari script - Hindi, Sanskrit, Marathi, Nepali, etc. | | JAPANESE | Japanese script - Hiragana, Katakana, Kanji | | KOREAN | Korean script - Hangul |

RecognitionLevel Enum (iOS)

| Value | Description | | ---------- | -------------------------------------------------------- | | ACCURATE | Slower but more accurate recognition (default). | | FAST | Faster but less accurate recognition. |

getSupportedLanguages()

Returns the list of supported languages on the current platform.

import { getSupportedLanguages } from '@zhanziyang/expo-text-extractor';

const languages = await getSupportedLanguages();
console.log(languages);
// iOS: ['en-US', 'fr-FR', 'de-DE', 'zh-Hans', 'zh-Hant', 'ja-JP', 'ko-KR', ...]
// Android: ['latin', 'chinese', 'devanagari', 'japanese', 'korean']

Platform-Specific Notes

Android (ML Kit)

• Each script requires a separate ML Kit model that is downloaded on-demand by Google Play Services.
• The first recognition request for a new script may be slower as the model downloads.
• All script models are included by default. If you want to reduce app size, you can create a custom build configuration.

iOS (Vision)

• The Vision framework automatically detects the script in most cases.
• Use the languages option for best results when you know the expected language(s).
• Call getSupportedLanguages() to get the full list of available languages for the device.
• Language support may vary by iOS version. Korean support was added in iOS 16.

Examples

Recognizing Chinese Text

import { extractTextFromImage, TextRecognitionScript } from '@zhanziyang/expo-text-extractor';

const recognizeChineseText = async (imageUri: string) => {
  const texts = await extractTextFromImage(imageUri, {
    script: TextRecognitionScript.CHINESE,
    // On iOS, you can also specify:
    // languages: ['zh-Hans', 'zh-Hant'],
  });
  
  return texts.join('\n');
};

Recognizing Japanese Text

import { extractTextFromImage, TextRecognitionScript } from '@zhanziyang/expo-text-extractor';

const recognizeJapaneseText = async (imageUri: string) => {
  const texts = await extractTextFromImage(imageUri, {
    script: TextRecognitionScript.JAPANESE,
    // On iOS:
    // languages: ['ja-JP'],
  });
  
  return texts.join('\n');
};

Recognizing Korean Text

import { extractTextFromImage, TextRecognitionScript } from '@zhanziyang/expo-text-extractor';

const recognizeKoreanText = async (imageUri: string) => {
  const texts = await extractTextFromImage(imageUri, {
    script: TextRecognitionScript.KOREAN,
    // On iOS:
    // languages: ['ko-KR'],
  });
  
  return texts.join('\n');
};

Multi-language Recognition (iOS)

On iOS, you can specify multiple preferred languages:

import { extractTextFromImage } from '@zhanziyang/expo-text-extractor';

const recognizeMultiLanguage = async (imageUri: string) => {
  const texts = await extractTextFromImage(imageUri, {
    // Recognize Japanese with English as fallback
    languages: ['ja-JP', 'en-US'],
  });
  
  return texts.join('\n');
};

Advanced iOS Options

Use all available iOS Vision options for fine-tuned recognition:

import { extractTextFromImage, RecognitionLevel } from '@zhanziyang/expo-text-extractor';

const advancedRecognition = async (imageUri: string) => {
  const texts = await extractTextFromImage(imageUri, {
    // Specify preferred languages
    languages: ['en-US'],
    
    // Or let iOS auto-detect (iOS 16+)
    automaticallyDetectsLanguage: true,
    
    // Enable language correction for better accuracy
    usesLanguageCorrection: true,
    
    // Add custom words for better recognition of proper nouns
    customWords: ['iPhone', 'MacBook', 'AirPods', 'WWDC'],
    
    // Ignore very small text (5% of image height)
    minimumTextHeight: 0.05,
    
    // Use accurate mode (default) or fast mode
    recognitionLevel: RecognitionLevel.ACCURATE,
  });
  
  return texts.join('\n');
};

Fast Recognition Mode (iOS)

For real-time processing or when speed is critical:

import { extractTextFromImage, RecognitionLevel } from '@zhanziyang/expo-text-extractor';

const fastRecognition = async (imageUri: string) => {
  const texts = await extractTextFromImage(imageUri, {
    recognitionLevel: RecognitionLevel.FAST,
    usesLanguageCorrection: false, // Disable for extra speed
  });
  
  return texts.join('\n');
};

Credits

This project is a fork of expo-text-extractor by pchalupa. Original work © pchalupa.

License

MIT