Edit

Share via

SpeechToText

  • Article

The SpeechToText API provides the ability to convert speech to text using online recognition. For offline recognition, you can use the OfflineSpeechToText.

Screenshot of SpeechText implemented on macOS

The following preconditions required for the SpeechToText:

Add permissions to AndroidManifest.xml:

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

For OfflineSpeechToText, Android 33.0 or higher is required.

Syntax

C#

The SpeechToText can be used as follows in C#:

async Task StartListening(CancellationToken cancellationToken)
{
    var isGranted = await speechToText.RequestPermissions(cancellationToken);
    if (!isGranted)
    {
        await Toast.Make("Permission not granted").Show(CancellationToken.None);
        return;
    }

    speechToText.RecognitionResultUpdated += OnRecognitionTextUpdated;
    speechToText.RecognitionResultCompleted += OnRecognitionTextCompleted;
    await speechToText.StartListenAsync(new SpeechToTextOptions	{ Culture = CultureInfo.CurrentCulture, ShouldReportPartialResults = true }, CancellationToken.None);
}

async Task StopListening(CancellationToken cancellationToken)
{
    await speechToText.StopListenAsync(CancellationToken.None);
    speechToText.RecognitionResultUpdated -= OnRecognitionTextUpdated;
    speechToText.RecognitionResultCompleted -= OnRecognitionTextCompleted;
}

void OnRecognitionTextUpdated(object? sender, SpeechToTextRecognitionResultUpdatedEventArgs args)
{
    RecognitionText += args.RecognitionResult;
}

void OnRecognitionTextCompleted(object? sender, SpeechToTextRecognitionResultCompletedEventArgs args)
{
    RecognitionText = args.RecognitionResult;
}

Methods

Method Description
RequestPermissions Asks for permission.
StartListenAsync Starts the SpeechToText service. (Real time speech recognition results will be surfaced via RecognitionResultUpdated and RecognitionResultCompleted)
StopListenAsync Stops the SpeechToText service. (Speech recognition results will be surfaced via RecognitionResultCompleted)

Properties

Property Type Description
CurrentState SpeechToTextState Gets a current listening state.

Events

EventName EventArgs Description
RecognitionResultUpdated SpeechToTextRecognitionResultUpdatedEventArgs Triggers when SpeechToText has real time updates.
RecognitionResultCompleted SpeechToTextRecognitionResultCompletedEventArgs Triggers when SpeechToText has completed.
StateChanged SpeechToTextStateChangedEventArgs Triggers when CurrentState has changed.

SpeechToTextOptions

The SpeechToTextOptions class provides the ability to configure the speech recognition service.

Properties

Property Type Description
Culture CultureInfo The spoken language to use for speech recognition.
ShouldReportPartialResults bool Gets or sets if include partial results. True by default.

SpeechToTextResult

The result returned from the RecognitionResultCompleted event. This can be used to verify whether the recognition was successful, and also access any exceptions that may have ocurred during the speech recognition.

Properties

Property Type Description
Text string The recognized text.
Exception Exception Gets the Exception if the speech recognition operation failed.
IsSuccessful bool Gets a value determining whether the operation was successful.

Methods

Method Description
EnsureSuccess Verifies whether the speech to text operation was successful.

Warning

EnsureSuccess will throw an Exception if the recognition operation was unsuccessful.

Dependency Registration

In case you want to inject service, you first need to register it. Update MauiProgram.cs with the next changes:

public static class MauiProgram
{
    public static MauiApp CreateMauiApp()
    {
        var builder = MauiApp.CreateBuilder();
        builder
            .UseMauiApp<App>()
			.UseMauiCommunityToolkit();

        builder.Services.AddSingleton<ISpeechToText>(SpeechToText.Default);
        // For offline recognition
        // builder.Services.AddSingleton<IOfflineSpeechToText>(OfflineSpeechToText.Default);
        return builder.Build();
    }
}

In case you need to register both SpeechToText and OfflineSpeechToText, you can use KeyedService.

Now you can inject the service like this:

public partial class MainPage : ContentPage
{
    private readonly ISpeechToText speechToText;

	public MainPage(ISpeechToText speechToText)
	{
		InitializeComponent();
        this.speechToText = speechToText;
	}
	
	public async void Listen(object sender, EventArgs args)
	{
		var isGranted = await speechToText.RequestPermissions(cancellationToken);
        if (!isGranted)
        {
            await Toast.Make("Permission not granted").Show(CancellationToken.None);
            return;
        }

        await speechToText.StartListenAsync(new SpeechToTextOptions { Culture = CultureInfo.CurrentCulture, ShouldReportPartialResults = true }, CancellationToken.None);
	}
}

Examples

You can find an example of SpeechToText in action in the .NET MAUI Community Toolkit Sample Application.

For offline recognition, you can use this sample: .NET MAUI Community Toolkit Sample Application.

API

You can find the source code for SpeechToText over on the .NET MAUI Community Toolkit GitHub repository.