azure_speech_recognition_services #

A Flutter plugin for Microsoft Azure Speech-to-Text that supports:

• single-shot recognition (listen once, stop on silence),
• continuous recognition (toggle start/stop stream), and
• optional pronunciation assessment callbacks.

What this plugin does #

azure_speech_recognition_services wraps native Azure Speech SDK behavior behind a Flutter MethodChannel API so you can:

• initialize Azure credentials once,
• register Dart callbacks for recognition events,
• start simple or continuous speech recognition, and
• receive partial/final transcription results in real time.

Core Dart API lives in AzureSpeechRecognitionServices.

Installation #

Add the package to your Flutter app:

dependencies:
  azure_speech_recognition_services: ^0.0.2

Then run:

flutter pub get

Azure prerequisites #

Before using the plugin, create an Azure Speech resource and get:

• subscriptionKey
• region (for example: westeurope, eastus, ...)

You pass those values to AzureSpeechRecognitionServices.initialize(...).

Platform notes #

This plugin uses native speech recognition integrations on mobile platforms.

• You should request microphone permission in your application.
• Use a valid Azure Speech subscription key + region.
• Language defaults to en-EN; pass lang explicitly if needed (for example en-US, it-IT, de-DE).

Complete usage example #

The following widget shows a full, practical setup for:

• initialization,
• callback wiring,
• single-shot recognition,
• continuous recognition toggle,
• cleanup.

import 'package:azure_speech_recognition_services/azure_speech_recognition_services.dart';
import 'package:flutter/material.dart';

class AzureSpeechDemo extends StatefulWidget {
  const AzureSpeechDemo({super.key});

  @override
  State<AzureSpeechDemo> createState() => _AzureSpeechDemoState();
}

class _AzureSpeechDemoState extends State<AzureSpeechDemo> {
  final AzureSpeechRecognitionServices _speech =
      AzureSpeechRecognitionServices();

  // Replace with your real values (or load from secure/env config).
  final String _subscriptionKey = 'YOUR_AZURE_SPEECH_KEY';
  final String _region = 'YOUR_AZURE_REGION';

  String _partialText = '';
  String _finalText = '';
  String _assessmentJson = '';
  String _status = 'Idle';
  bool _isContinuous = false;

  @override
  void initState() {
    super.initState();

    AzureSpeechRecognitionServices.initialize(
      _subscriptionKey,
      _region,
      lang: 'en-US',
      timeout: '2000', // valid range: 100..5000
    );

    _speech.setRecognitionStartedHandler(() {
      setState(() => _status = 'Recognition started');
    });

    _speech.setRecognitionResultHandler((text) {
      setState(() => _partialText = text);
    });

    _speech.setFinalTranscription((text) {
      setState(() {
        _finalText = text;
        _status = 'Final transcription received';
      });
    });

    _speech.setAssessmentResult((json) {
      setState(() => _assessmentJson = json);
    });

    _speech.setStartHandler(() {
      setState(() {
        _isContinuous = true;
        _status = 'Continuous recognition ON';
      });
    });

    _speech.setRecognitionStoppedHandler(() {
      setState(() {
        _isContinuous = false;
        _status = 'Continuous recognition OFF';
      });
    });

    _speech.onExceptionHandler((message) {
      setState(() => _status = 'Error: $message');
    });
  }

  Future<void> _startSingleShot() async {
    AzureSpeechRecognitionServices.simpleVoiceRecognition();
  }

  Future<void> _toggleContinuous() async {
    AzureSpeechRecognitionServices.continuousRecording();
  }

  Future<void> _singleShotWithAssessment() async {
    AzureSpeechRecognitionServices.simpleVoiceRecognitionWithAssessment(
      referenceText: 'The quick brown fox jumps over the lazy dog.',
      granularity: 'Phoneme',
      phonemeAlphabet: 'IPA',
      enableMiscue: true,
      nBestPhonemeCount: 5,
    );
  }

  @override
  void dispose() {
    // Defensive cleanup if continuous recognition is still active.
    AzureSpeechRecognitionServices.stopContinuousRecognition();
    super.dispose();
  }

  @override
  Widget build(BuildContext context) {
    return Scaffold(
      appBar: AppBar(title: const Text('Azure Speech Demo')),
      body: Padding(
        padding: const EdgeInsets.all(16),
        child: Column(
          crossAxisAlignment: CrossAxisAlignment.start,
          children: [
            Text('Status: $_status'),
            const SizedBox(height: 12),
            Text('Partial: $_partialText'),
            const SizedBox(height: 8),
            Text('Final: $_finalText'),
            const SizedBox(height: 8),
            Text('Assessment JSON: $_assessmentJson'),
            const Spacer(),
            Wrap(
              spacing: 12,
              runSpacing: 12,
              children: [
                ElevatedButton(
                  onPressed: _startSingleShot,
                  child: const Text('Single-shot recognize'),
                ),
                ElevatedButton(
                  onPressed: _singleShotWithAssessment,
                  child: const Text('Single-shot + assessment'),
                ),
                ElevatedButton(
                  onPressed: _toggleContinuous,
                  child: Text(
                    _isContinuous
                        ? 'Stop continuous'
                        : 'Start continuous',
                  ),
                ),
              ],
            ),
          ],
        ),
      ),
    );
  }
}

API quick reference #

1) Initialization #

AzureSpeechRecognitionServices.initialize(
  subscriptionKey,
  region,
  lang: 'en-US',
  timeout: '2000',
);

• timeout must be an integer string in range 100..5000.

2) Event handlers #

Register handlers on the singleton instance:

• setRecognitionStartedHandler
• setRecognitionResultHandler (partial text)
• setFinalTranscription (final text)
• setAssessmentResult (assessment JSON/text)
• setStartHandler (continuous mode started)
• setRecognitionStoppedHandler (continuous mode stopped)
• onExceptionHandler

3) Recognition methods #

• simpleVoiceRecognition()
• simpleVoiceRecognitionWithAssessment(...)
• continuousRecording() (toggle)
• continuousRecordingWithAssessment(...) (toggle)
• isContinuousRecognitionOn()
• stopContinuousRecognition()

Recommended integration pattern #

1. Create one AzureSpeechRecognitionServices() instance in your state object.
2. Call initialize(...) once (for current credentials/language).
3. Register all callbacks in initState.
4. Trigger recognition from explicit UI actions (buttons/gestures).
5. Stop continuous recognition in dispose for safe cleanup.

Formatting #

On Windows, dart format . can fail when generated Android build artifacts contain invalid or missing transformed directories.

Use the repo formatter scripts instead:

./tool/format.sh

./tool/format.ps1

These scripts format only tracked Dart source files, so generated directories like build/ and example/build/ are automatically ignored.