feat: add Speech Recognition sample and enable Phi on GPU

AIDevGallery/Controls/WcrModelDownloader.xaml

@@ -31,7 +31,7 @@
                 IsTextSelectionEnabled="True"
                 TextAlignment="Center"
                 TextWrapping="Wrap">
-                <Run Text="This Windows AI API requires a one-time model download via Windows Update." /><LineBreak /> <LineBreak />
+                <Run x:Name="ModelDownloadInfoRun" Text="This Windows AI API requires a one-time model download via Windows Update." /><LineBreak /> <LineBreak />
                 <Hyperlink NavigateUri="https://learn.microsoft.com/windows/ai/apis/model-setup#prerequisites" UnderlineStyle="None">A Copilot+ PC with Windows 11 Build 26120.3073 or higher is required</Hyperlink>
             </TextBlock>
             <TextBlock
@@ -71,7 +71,15 @@
                 TextWrapping="WrapWholeWords">
                 <Run Text="Requesting model.." /> <LineBreak />
                 <LineBreak />
-                <Run FontSize="12" Text="Download progress can also be tracked in " /><Hyperlink Click="WindowsUpdateHyperlinkClicked" FontSize="12">Windows Update</Hyperlink><LineBreak />
+            </TextBlock>
+            <TextBlock
+                x:Name="WindowsUpdateTrackingText"
+                Margin="0,0,0,24"
+                HorizontalAlignment="Center"
+                Foreground="{ThemeResource TextFillColorSecondaryBrush}"
+                TextAlignment="Center"
+                TextWrapping="WrapWholeWords">
+                <Run FontSize="12" Text="Download progress can also be tracked in " /><Hyperlink Click="WindowsUpdateHyperlinkClicked" FontSize="12">Windows Update</Hyperlink>
             </TextBlock>
         </StackPanel>
 

AIDevGallery/Controls/WcrModelDownloader.xaml.cs

@@ -166,6 +166,15 @@ public Task<bool> SetDownloadOperation(ModelType modelType, string sampleId, Fun
             ? Visibility.Visible
             : Visibility.Collapsed;
 
+        // TODO: Remove once the Speech Recognition ships through Windows Update
+        var isSpeechRecognition = modelType == ModelType.SpeechRecognition;
+        ModelDownloadInfoRun.Text = isSpeechRecognition
+            ? "This Windows AI API requires a one-time model download."
+            : "This Windows AI API requires a one-time model download via Windows Update.";
+        WindowsUpdateTrackingText.Visibility = isSpeechRecognition
+            ? Visibility.Collapsed
+            : Visibility.Visible;
+
         if (exisitingOperation != null && exisitingOperation.Status == AsyncStatus.Started)
         {
             // don't reuse same one because we can only have one Progress delegate

AIDevGallery/Package.Store.appxmanifest

@@ -62,5 +62,6 @@
   <Capabilities>
     <rescap:Capability Name="runFullTrust" />
 	<systemai:Capability Name="systemAIModels"/>
+    <DeviceCapability Name="microphone" />
   </Capabilities>
 </Package>

AIDevGallery/Package.appxmanifest

@@ -62,5 +62,6 @@
   <Capabilities>
     <rescap:Capability Name="runFullTrust" />
     <systemai:Capability Name="systemAIModels"/>
+    <DeviceCapability Name="microphone" />
   </Capabilities>
 </Package>

AIDevGallery/Samples/Definitions/WcrApis/WcrApiCodeSnippet.cs

@@ -721,6 +721,65 @@ public async void IndexStatisticsSample()
                 }
             }
             """""
+        },
+        {
+            ModelType.SpeechRecognition, """"
+            using Microsoft.Windows.AI;
+            using Microsoft.Windows.AI.MachineLearning;
+            using Microsoft.Windows.AI.Speech;
+
+            // Speech recognition runs on the device's execution providers (CPU / GPU / NPU).
+            var catalog = ExecutionProviderCatalog.GetDefault();
+            await catalog.EnsureAndRegisterCertifiedAsync();
+
+            var readyState = SpeechRecognitionModel.GetReadyState();
+            if (readyState is AIFeatureReadyState.Ready or AIFeatureReadyState.NotReady)
+            {
+                if (readyState == AIFeatureReadyState.NotReady)
+                {
+                    var ensureOp = await SpeechRecognitionModel.EnsureReadyAsync();
+                    if (ensureOp.Status != AIFeatureReadyResultState.Success)
+                    {
+                        throw new InvalidOperationException("Speech model could not be prepared.");
+                    }
+                }
+
+                var modelResult = await SpeechRecognitionModel.TryCreateAsync();
+                if (modelResult.ExtendedError != null)
+                {
+                    throw modelResult.ExtendedError;
+                }
+
+                using SpeechRecognitionModel speechModel = modelResult.SpeechModel;
+
+                // Stream audio from the default microphone. Pass an empty deviceId to use the system default.
+                var audioConfig = AudioConfiguration.FromAudioDevice(string.Empty);
+
+                using var streaming = new StreamingRecognition(audioConfig, speechModel);
+
+                streaming.Recognizing += (_, args) =>
+                {
+                    // Interim hypothesis (updates frequently as more audio arrives).
+                    Console.WriteLine($"[interim] {args.Text}");
+                };
+
+                streaming.Recognized += (_, args) =>
+                {
+                    // Final result for a stable utterance.
+                    Console.WriteLine($"[final] offset={args.Offset:F2}s duration={args.Duration:F2}s: {args.Text}");
+                };
+
+                await streaming.StartContinuousRecognitionAsync();
+
+                // ... let captions stream in. When done:
+                streaming.StopContinuousRecognition();
+            }
+            else
+            {
+                // DisabledByUser or NotSupportedOnCurrentSystem.
+                throw new InvalidOperationException($"Speech recognition is not available: {readyState}.");
+            }
+            """"
         }
     };
 }

AIDevGallery/Samples/Definitions/WcrApis/WcrApiHelpers.cs

@@ -5,10 +5,13 @@
 using AIDevGallery.Utils;
 using Microsoft.Windows.AI;
 using Microsoft.Windows.AI.Imaging;
+using Microsoft.Windows.AI.MachineLearning;
+using Microsoft.Windows.AI.Speech;
 using Microsoft.Windows.AI.Text;
 using Microsoft.Windows.AI.Video;
 using System;
 using System.Collections.Generic;
+using System.Runtime.InteropServices.WindowsRuntime;
 using Windows.Foundation;
 
 namespace AIDevGallery.Samples;
@@ -80,6 +83,9 @@ internal static class WcrApiHelpers
         },
         {
             ModelType.VideoSuperRes, VideoScaler.GetReadyState
+        },
+        {
+            ModelType.SpeechRecognition, SpeechRecognitionModel.GetReadyState
         }
     };
 
@@ -132,9 +138,29 @@ internal static class WcrApiHelpers
         },
         {
             ModelType.VideoSuperRes, VideoScaler.EnsureReadyAsync
+        },
+        {
+            ModelType.SpeechRecognition, EnsureSpeechRecognitionModelReadyAsync
         }
     };
 
+    // SpeechRecognitionModel.EnsureReadyAsync reports progress as SpeechRecognitionModelProgress,
+    // so adapt it to the IAsyncOperationWithProgress<AIFeatureReadyResult, double> shape the gallery expects.
+    private static IAsyncOperationWithProgress<AIFeatureReadyResult, double> EnsureSpeechRecognitionModelReadyAsync()
+    {
+        return AsyncInfo.Run<AIFeatureReadyResult, double>(async (cancellationToken, progress) =>
+        {
+            progress.Report(0);
+            var catalog = ExecutionProviderCatalog.GetDefault();
+            await catalog.EnsureAndRegisterCertifiedAsync().AsTask(cancellationToken);
+
+            var inner = SpeechRecognitionModel.EnsureReadyAsync();
+            inner.Progress = (_, p) => progress.Report(p.Progress);
+            using var registration = cancellationToken.Register(() => inner.Cancel());
+            return await inner;
+        });
+    }
+
     // this is a workaround for GetReadyState not returning Ready after EnsureReadyAsync is called
     // for now, we will track when EnsureReadyAsync succeeds for each model to ensure we are not
     // blocking the samples from running until this bug is fixed

AIDevGallery/Samples/Definitions/WcrApis/apis.json

@@ -218,6 +218,17 @@
         "ReadmeUrl": "https://github.com/MicrosoftDocs/windows-ai-docs/blob/docs/docs/apis/video-super-resolution.md",
         "License": "ms-pl",
         "SampleIdToShowInDocs": "c3252e18-1d47-4689-adae-78fc66968650"
+      },
+      "SpeechRecognition": {
+        "Id": "0d4f1c2a-7e3b-4c9e-9b8a-1f2d3c4b5a60",
+        "Name": "Speech Recognition",
+        "Icon": "WCRAPI.svg",
+        "IconGlyph": "\uE720",
+        "Description": "Continuously transcribe audio using local speech recognition.",
+        "ReadmeUrl": "https://github.com/MicrosoftDocs/windows-ai-docs/blob/docs/docs/apis/speech-recognition.md",
+        "License": "ms-pl",
+        "SampleIdToShowInDocs": "9c5b2e8a-1f7d-4d3c-9e6a-3b1c8e7f4d20",
+        "Category": "Speech"
       }
     }
   }

AIDevGallery/Samples/WCRAPIs/SpeechRecognition.xaml

@@ -0,0 +1,97 @@
+<?xml version="1.0" encoding="utf-8" ?>
+<samples:BaseSamplePage
+    x:Class="AIDevGallery.Samples.WCRAPIs.SpeechRecognition"
+    xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
+    xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
+    xmlns:d="http://schemas.microsoft.com/expression/blend/2008"
+    xmlns:mc="http://schemas.openxmlformats.org/markup-compatibility/2006"
+    xmlns:samples="using:AIDevGallery.Samples"
+    mc:Ignorable="d">
+    <Grid RowSpacing="12">
+        <Grid.RowDefinitions>
+            <RowDefinition Height="*" />
+            <RowDefinition Height="Auto" />
+            <RowDefinition Height="Auto" />
+        </Grid.RowDefinitions>
+
+        <Border
+            Grid.Row="0"
+            Background="{ThemeResource CardBackgroundFillColorDefaultBrush}"
+            BorderBrush="{ThemeResource ControlStrongStrokeColorDefaultBrush}"
+            BorderThickness="1"
+            CornerRadius="{StaticResource OverlayCornerRadius}">
+            <ScrollViewer
+                x:Name="TranscriptionScrollViewer"
+                Padding="16"
+                HorizontalScrollBarVisibility="Disabled"
+                VerticalScrollBarVisibility="Auto">
+                <StackPanel Spacing="4">
+                    <TextBlock
+                        x:Name="FinalTranscriptionTextBlock"
+                        AutomationProperties.LiveSetting="Polite"
+                        AutomationProperties.Name="Final transcription"
+                        FontFamily="Cascadia Code"
+                        FontSize="16"
+                        IsTextSelectionEnabled="True"
+                        Text="Press Start to begin speech recognition..."
+                        TextWrapping="Wrap" />
+                    <TextBlock
+                        x:Name="InterimTranscriptionTextBlock"
+                        AutomationProperties.LiveSetting="Polite"
+                        AutomationProperties.Name="Interim transcription"
+                        FontFamily="Cascadia Code"
+                        FontSize="16"
+                        FontStyle="Italic"
+                        Foreground="{ThemeResource TextFillColorSecondaryBrush}"
+                        IsTextSelectionEnabled="True"
+                        TextWrapping="Wrap" />
+                </StackPanel>
+            </ScrollViewer>
+        </Border>
+
+        <InfoBar
+            x:Name="StatusInfoBar"
+            Grid.Row="1"
+            IsClosable="False"
+            IsOpen="False"
+            Severity="Informational" />
+
+        <StackPanel
+            Grid.Row="2"
+            HorizontalAlignment="Center"
+            Orientation="Horizontal"
+            Spacing="8">
+            <ComboBox
+                x:Name="InputSourceComboBox"
+                MinWidth="240"
+                AutomationProperties.Name="Input source"
+                SelectedIndex="0"
+                ToolTipService.ToolTip="Choose what to transcribe, then press Start.">
+                <ComboBoxItem
+                    Content="Microphone (live)"
+                    Tag="Microphone"
+                    ToolTipService.ToolTip="Live transcription from the default microphone using StreamingRecognition." />
+                <ComboBoxItem
+                    Content="Audio file - full transcript"
+                    Tag="FileBatch"
+                    ToolTipService.ToolTip="BatchRecognition.RecognizeFromFile: one-shot call that returns the complete transcript." />
+                <ComboBoxItem
+                    Content="Audio file - incremental"
+                    Tag="FileStreaming"
+                    ToolTipService.ToolTip="StreamingRecognition over a file: raises incremental Recognizing/Recognized events, like the microphone." />
+            </ComboBox>
+            <Button
+                x:Name="StartStopButton"
+                MinWidth="160"
+                Click="StartStopButton_Click"
+                Content="Start recognition"
+                Style="{StaticResource AccentButtonStyle}" />
+            <Button
+                x:Name="ClearButton"
+                MinWidth="120"
+                AutomationProperties.Name="Clear transcription"
+                Click="ClearButton_Click"
+                Content="Clear" />
+        </StackPanel>
+    </Grid>
+</samples:BaseSamplePage>