Vosk Flutter Service #

Flutter plugin for offline speech recognition using the Vosk toolkit.

Acknowledgement: This package is a modernized fork of the original vosk_flutter plugin, updated to support the latest Flutter/Dart SDKs, maintain strict type safety, and expand platform support (iOS/macOS).

Platform Support #

Note: Web support is currently a stub for platform compatibility. Full WASM implementation is in progress.

Installation #

Dependencies #

Add vosk_flutter_service to your pubspec.yaml:

dependencies:
  vosk_flutter_service: ^0.0.1+1

iOS Configuration #

1. Framework: You must explicitly provide the LibVosk.xcframework.
   • Download it from the official Vosk releases or build it.
   • Place LibVosk.xcframework inside ios/Frameworks.
2. Permissions: Add the microphone usage description to Info.plist:

   <key>NSMicrophoneUsageDescription</key>
   <string>This app needs access to the microphone for speech recognition.</string>
   

Android Configuration #

Add ProGuard rules to android/app/proguard-rules.pro:

-keep class com.sun.jna.* { *; }
-keepclassmembers class * extends com.sun.jna.* { public *; }

Add microphone permission to AndroidManifest.xml:

<uses-permission android:name="android.permission.RECORD_AUDIO" />

Usage #

1. Load Model #

Load the model from assets or storage.

final vosk = VoskFlutterPlugin(); // Use default constructor
final modelPath = await ModelLoader().loadFromAssets('assets/models/vosk-model-small-en-us-0.15.zip');
final model = await vosk.createModel(modelPath);

2. Create Recognizer #

final recognizer = await vosk.createRecognizer(
  model: model,
  sampleRate: 16000,
);

3. Recognize Audio #

Feed audio data (PCM 16-bit mono) to the recognizer:

await recognizer.acceptWaveformBytes(audioBytes);
print(await recognizer.getFinalResult());

4. Continuous Microphone Recognition (Android/iOS) #

final speechService = await vosk.initSpeechService(recognizer);
await speechService.start();

speechService.onPartial().listen((partial) => print(partial));
speechService.onResult().listen((result) => print(result));

For Desktop (Linux/Windows/macOS), use a separate package (like record or flutter_sound) to capture audio stream and feed it to acceptWaveformBytes.