Speech Recognition in Uno

[!TIP] This article covers Uno-specific information for Windows.Media.SpeechRecognition. For a full description of the feature and instructions on using it, see SpeechRecognition Namespace.

The Windows.Media.SpeechRecognition class allows an application to recognize voice input.

Uno's implementation currently supports basic native speech recognition.

Supported Features

The following features of Windows.Media.SpeechRecognition.SpeechRecognizer are currently supported:

Feature	iOS	Android	macOS (Skia)	WebAssembly	Remarks
SpeechRecognizer()	X	X	X	X
SpeechRecognizer(Language)	X	X	X	X
Constraints	-	-	-	-
ContinuousRecognitionSession	-	-	-	-
CurrentLanguage	X	X	X	X
State	X	X	X	X
SupportedGrammarLanguages	-	-	-	-
SupportedTopicLanguages	-	-	-	-
SystemSpeechLanguage	-	-	-	-
Timeouts	X	X	X	X
UIOptions	X	X	X	X	Not used
CompileConstraintsAsync()	X	X	X	X	Always return Success (implemented to meet WinUI constraint that requires `CompileConstraintsAsync()` to be called before `RecognizeAsync()`)
Dispose()	X	X	X	X
RecognizeAsync()	X	X	X	X
RecognizeWithUIAsync()	-	-	-	-
StopRecognitionAsync()	X	X	X	X
TrySetSystemSpeechLanguageAsync(Language)	-	-	-	-
HypothesisGenerated	X	X	X	X
RecognitionQualityDegrading	-	-	-	-
StateChanged	X	X	X	X

Requirement

iOS

iOS 10 or later is required.

The following lines need to be added to your info.plist:

xml

<key>NSSpeechRecognitionUsageDescription</key>  
<string>[SpeechRecognition usage description]</string>  
<key>NSMicrophoneUsageDescription</key>  
<string>[SpeechRecognition usage description]</string>

Android

The following lines need to be added to your AndroidManifest.xml:

xml

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

macOS (Skia)

macOS 10.15 or later is required.

The following lines need to be added to your app bundle's Info.plist:

xml

<key>NSSpeechRecognitionUsageDescription</key>
<string>[SpeechRecognition usage description]</string>
<key>NSMicrophoneUsageDescription</key>
<string>[SpeechRecognition usage description]</string>

The app must be packaged as an .app bundle when running speech recognition — macOS only displays the TCC authorization prompt for bundled apps. Launching the executable directly from a terminal (dotnet run) will result in SFSpeechRecognizer returning Denied without prompting.

[!NOTE] For contributors: The Uno samples app (SamplesApp.Skia.Generic) ships a build target that automatically produces a minimal .app bundle next to the regular output on macOS. After dotnet build, launch it with:
bash
open src/SamplesApp/SamplesApp.Skia.Generic/bin/Debug/net10.0/SamplesApp.Skia.Generic.app

WebAssembly

Speech recognition on WebAssembly relies on the browser's Web Speech API implementation:

window.SpeechRecognition, or
window.webkitSpeechRecognition (vendor-prefixed fallback).

Requirements and behavior notes:

The app must run in a browser that exposes one of the APIs above.
The user must grant microphone permission in the browser.
Browsers typically require a secure context (https://), with http://localhost commonly allowed for local development.
Starting recognition may require user interaction (for example, a button click) depending on browser policies.

Limitation

In Windows.Media.SpeechRecognition.SpeechRecognitionResult, only Text, Alternates, and GetAlternates(uint maxAlternates) are implemented. In particular, RawConfidence and Confidence fields are not currently supported.

On WebAssembly specifically, availability is browser-dependent. If speech recognition is not available in the current browser, RecognizeAsync() fails with an InvalidOperationException.