Vosk Flutter Plugin

Flutter plugin for Vosk speech recognition.

Platform Support

Android	iOS	Web	Linux	Windows
✔	✔	➖	✔	✔

Usage

Native Binaries Installation

This plugin requires native Vosk binaries to function on most platforms. Follow the instructions below to install them:

Android

Native binaries are included in the package. No additional steps are required.

iOS, Linux, and Windows

Run the following command in your project root to download and install the native binaries:

dart run vosk_flutter_service install -t <platform>

Replace <platform> with ios, linux, or windows.

Configurations

Follow the instruction at the Installing page of the package.

Android

Add this pro guard rules in android/app/proguard-rules.pro(if the file does not exist - create it):

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

If you want to use a microphone input, add the microphone permission to your AndroidManifest.xml:

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

Load model

flutter:
  assets:
    - assets/models/

final vosk = VoskFlutterPlugin.instance();
final enSmallModelPath = await ModelLoader()
    .loadFromAssets('assets/models/vosk-model-small-en-us-0.15.zip');

Create recognizer

final recognizer = await vosk.createRecognizer(
    model: model,
    sampleRate: sampleRate,
);
final recognizerWithGrammar = await vosk.createRecognizer(
    model: model,
    sampleRate: sampleRate,
    grammar: ['one', 'two', 'three'],
);

Recognize audio data

Uint8List audioBytes = ...; // audio data in PCM 16-bit mono format
List<String> results = [];
int chunkSize = 8192;
int pos = 0;

while (pos + chunkSize < audioBytes.length) {
    final resultReady = await recognizer.acceptWaveformBytes(
      Uint8List.fromList(audioBytes.getRange(pos, pos + chunkSize).toList()));
    pos += chunkSize;
    
    if (resultReady) {
      print(await recognizer.getResult());
    } else {
      print(await recognizer.getPartialResult());
    }
}
await recognizer.acceptWaveformBytes(
  Uint8List.fromList(audioBytes.getRange(pos, audioBytes.length).toList()));
print(await recognizer.getFinalResult());

Recognize microphone data

Android

final speechService = await vosk.initSpeechService(recognizer);
speechService.onPartial().forEach((partial) => print(partial));
speechService.onResult().forEach((result) => print(result));
await speechService.start();

Linux & Windows

Use any suitable plugin to get the microphone input and pass it to a recognizer