SpeechRecognition

Returns true when the runtime exposes a SpeechRecognition implementation - standard or webkit-prefixed. The Butil script resolves both, so you never have to feature-detect the prefix yourself.

@inject Bit.Butil.SpeechRecognition speechRecognition

var isSupported = await speechRecognition.IsSupported();

Start begins listening and returns an IAsyncDisposable handle. Options select the language, whether recognition keeps running across pauses (Continuous), whether interim non-final transcripts are reported (InterimResults) and how many alternatives to surface. Each result arrives in the onResult callback as a SpeechRecognitionResult with Transcript, Confidence and IsFinal.

await using var session = await speechRecognition.Start(
    new SpeechRecognitionOptions
    {
        Lang = "en-US",
        Continuous = true,
        InterimResults = true,
        MaxAlternatives = 1,
    },
    onResult: result =>
    {
        // result.Transcript, result.Confidence (0-1), result.IsFinal
    },
    onError: message => { /* mic denied, no speech, network, ... */ },
    onEnd: () => { /* engine stopped listening */ });

Disposing the handle returned by Start stops that session and detaches its callbacks - an await using block covers the common case. The service itself is IAsyncDisposable too and tears down every live session when the scope ends, but pages that hold a handle across renders should dispose it explicitly in their own DisposeAsync.

@implements IAsyncDisposable

@code {
    private IAsyncDisposable? session;

    private async Task Stop()
    {
        if (session is null) return;
        await session.DisposeAsync(); // stops the engine and detaches callbacks
        session = null;
    }

    public async ValueTask DisposeAsync() => await Stop();
}