Masamune Google Speech-to-Text

[GitHub] | [YouTube] | [Packages] | [X] | [LinkedIn] | [mathru.net]

Masamune Google Speech-to-Text #

Usage #

Installation #

Add the package to your project.

flutter pub add masamune_speech_to_text

Run flutter pub get when editing pubspec.yaml manually.

Register the Adapter #

Configure GoogleSpeechToTextMasamuneAdapter before launching the app. Provide the default language settings.

// lib/adapter.dart

/// Masamune adapters used in the application.
final masamuneAdapters = <MasamuneAdapter>[
  const UniversalMasamuneAdapter(),

  const GoogleSpeechToTextMasamuneAdapter(
    defaultLocale: Locale('en', 'US'),  // Default language (required)
  ),
];

For Japanese:

const GoogleSpeechToTextMasamuneAdapter(
  defaultLocale: Locale('ja', 'JP'),
)

With Auto-initialization (optional):

// Create controller for auto-initialization
final sttController = SpeechToTextController();

const GoogleSpeechToTextMasamuneAdapter(
  defaultLocale: Locale('en', 'US'),
  speechToTextController: sttController,  // Will initialize on boot
  initializeOnBoot: true,                 // Auto-initialize
)

Speech-to-Text Controller #

Use SpeechToTextController to initialize speech recognition, start listening, and handle results.

class VoiceInputPage extends PageScopedWidget {
  @override
  Widget build(BuildContext context, PageRef ref) {
    final stt = ref.page.controller(SpeechToTextController.query());

    // Initialize on page load
    ref.page.on(
      initOrUpdate: () {
        stt.initialize();
      },
    );

    return Scaffold(
      appBar: AppBar(title: const Text("Voice Input")),
      body: Column(
        children: [
          Text(stt.recognizedText ?? "Say something..."),
          
          ElevatedButton(
            onPressed: () async {
              if (stt.isListening) {
                await stt.stop();
              } else {
                await stt.listen(
                  onResult: (result) {
                    print("Recognized: ${result.recognizedWords}");
                  },
                );
              }
            },
            child: Text(stt.isListening ? "Stop" : "Start Listening"),
          ),
        ],
      ),
    );
  }
}

Continuous Listening #

• listen() starts recognition; set partialResults: true to receive interim transcripts.
• Use stt.pause() / stt.resume() to manage listening sessions without full reinitialization.
• Handle permission checks; the controller throws if microphone access is denied.

Error Handling #

Listen to the error stream or supply onError to listen() to capture SpeechRecognitionError details.

await stt.listen(
  onError: (SpeechRecognitionError error) {
    debugPrint("Error: ${error.errorMsg}");
  },
);

Tips #

• Always call initialize() before listening; reuse the controller across sessions for performance.
• Provide UI feedback (animations, status indicators) while listening.
• Localize locale IDs (e.g., ja_JP, fr_FR) to match target audiences.
• Combine with text input fields to let users edit recognized text.

GitHub Sponsors #

Sponsors are always welcome. Thank you for your support!

https://github.com/sponsors/mathrunet