Merge branch 'master' of https://github.com/microsoft/cognitive-services-speech-sdk-js into release/1.43

Author: Brian Mouncer

cspell.json

@@ -0,0 +1,5 @@
+{
+    "words": [
+        "viseme"
+    ]
+}

src/common.speech/AvatarSynthesisAdapter.ts

@@ -77,6 +77,7 @@ export class AvatarSynthesisAdapter extends SynthesisAdapterBase {
                 character: this.privAvatarConfig.character,
                 customized: this.privAvatarConfig.customized,
                 style: this.privAvatarConfig.style,
+                useBuiltInVoice: this.privAvatarConfig.useBuiltInVoice,
             }
         } as ISynthesisSectionVideo;
     }

src/common.speech/SpeechServiceConfig.ts

@@ -167,6 +167,7 @@ export interface ISynthesisSectionVideo {
     talkingAvatar: {
         character: string;
         customized: boolean;
+        useBuiltInVoice: boolean;
         style: string;
         background: {
             color: string;

src/sdk/AvatarConfig.ts

@@ -13,6 +13,7 @@ import { AvatarVideoFormat } from "./Exports.js";
  */
 export class AvatarConfig {
     private privCustomized: boolean = false;
+    private privUseBuiltInVoice: boolean = false;
     private privBackgroundColor: string;
     private privBackgroundImage: URL;
     private privRemoteIceServers: RTCIceServer[];
@@ -44,6 +45,20 @@ export class AvatarConfig {
         this.privCustomized = value;
     }
 
+    /**
+     * Indicates whether to use built-in voice for custom avatar.
+     */
+    public get useBuiltInVoice(): boolean {
+        return this.privUseBuiltInVoice;
+    }
+
+    /**
+     * Sets whether to use built-in voice for custom avatar.
+     */
+    public set useBuiltInVoice(value: boolean) {
+        this.privUseBuiltInVoice = value;
+    }
+
     /**
      * Gets the background color.
      */

src/sdk/Exports.ts

@@ -98,7 +98,7 @@ export { SpeechSynthesisBookmarkEventArgs } from "./SpeechSynthesisBookmarkEvent
 export { SpeechSynthesisVisemeEventArgs } from "./SpeechSynthesisVisemeEventArgs.js";
 export { SpeechSynthesisBoundaryType } from "./SpeechSynthesisBoundaryType.js";
 export { SynthesisVoicesResult } from "./SynthesisVoicesResult.js";
-export { VoiceInfo } from "./VoiceInfo.js";
+export { SynthesisVoiceGender, SynthesisVoiceType, VoiceInfo } from "./VoiceInfo.js";
 export { IPlayer } from "./Audio/IPlayer.js";
 export { SpeakerAudioDestination } from "./Audio/SpeakerAudioDestination.js";
 export { CancellationEventArgs } from "./CancellationEventArgs.js";

src/sdk/VoiceInfo.ts

@@ -23,10 +23,35 @@ export enum SynthesisVoiceGender {
 }
 
 export enum SynthesisVoiceType {
+    /**
+     * Voice type is not known.
+     */
+    Unknown = 0,
+
+    /**
+     * Online neural voices.
+     */
     OnlineNeural = 1,
+
+    /**
+     * Online standard voices. These voices are deprecated.
+     */
     OnlineStandard = 2,
+
+    /**
+     * Offline neural voices.
+     */
     OfflineNeural = 3,
+
+    /**
+     * Offline standard voices.
+     */
     OfflineStandard = 4,
+
+    /**
+     * High definition (HD) voices. Refer to https://learn.microsoft.com/azure/ai-services/speech-service/high-definition-voices
+     */
+    OnlineNeuralHD = 5,
 }
 
 const GENDER_LOOKUP: Record<string, SynthesisVoiceGender> = {
@@ -35,6 +60,11 @@ const GENDER_LOOKUP: Record<string, SynthesisVoiceGender> = {
     [SynthesisVoiceGender[SynthesisVoiceGender.Female]]: SynthesisVoiceGender.Female,
 };
 
+const VOICE_TYPE_LOOKUP: Record<string, SynthesisVoiceType> = {
+    Neural: SynthesisVoiceType.OnlineNeural,
+    NeuralHD: SynthesisVoiceType.OnlineNeuralHD,
+};
+
 /**
  * Information about Speech Synthesis voice
  * Added in version 1.20.0.
@@ -66,7 +96,7 @@ export class VoiceInfo {
             this.privLocaleName = json.LocaleName;
             this.privDisplayName = json.DisplayName;
             this.privLocalName = json.LocalName;
-            this.privVoiceType = json.VoiceType.endsWith("Standard") ? SynthesisVoiceType.OnlineStandard : SynthesisVoiceType.OnlineNeural;
+            this.privVoiceType = VOICE_TYPE_LOOKUP[json.VoiceType] || SynthesisVoiceType.Unknown;
             this.privGender = GENDER_LOOKUP[json.Gender] || SynthesisVoiceGender.Unknown;
 
             if (!!json.StyleList && Array.isArray(json.StyleList)) {

tests/SpeechSynthesisTests.ts

@@ -165,6 +165,16 @@ test("testGetVoicesAsyncDefault", async () => {
     expect(voicesResult.resultId).not.toBeUndefined();
     expect(voicesResult.voices.length).toBeGreaterThan(0);
     expect(voicesResult.reason).toEqual(sdk.ResultReason.VoicesListRetrieved);
+
+    for (const voice of voicesResult.voices) {
+        expect(voice.name).not.toBeUndefined();
+        expect(voice.locale).not.toBeUndefined();
+        expect(voice.shortName).not.toBeUndefined();
+        expect(voice.displayName).not.toBeUndefined();
+        expect(voice.localName).not.toBeUndefined();
+        expect(voice.gender).not.toEqual(sdk.SynthesisVoiceGender.Unknown);
+        expect(voice.voiceType).not.toEqual(sdk.SynthesisVoiceType.Unknown);
+    }
 });
 
 test("testGetVoicesAsyncAuthWithToken", async () => {
@@ -228,6 +238,8 @@ test("testGetVoicesAsyncUS", async () => {
     expect(ava).not.toBeUndefined();
     expect(ava.voiceTag.TailoredScenarios).not.toBeUndefined();
     expect(ava.voiceTag.TailoredScenarios[0]).toEqual("Chat");
+    expect(ava.gender).toEqual(sdk.SynthesisVoiceGender.Female);
+    expect(ava.voiceType).toEqual(sdk.SynthesisVoiceType.OnlineNeural);
 });
 
 Settings.testIfDOMCondition("testSpeechSynthesizer1", () => {