ssml-builder-js 2.7.0 → 2.8.1

package/dist/elements.js

@@ -74,7 +74,8 @@ var SSML_TAGS = {
   MSTTS_SILENCE: "mstts:silence",
   SILENCE: "silence",
   MSTTS_VISEME: "mstts:viseme",
-  VISEME: "viseme"
+  VISEME: "viseme",
+  MSTTS_AUDIO_DURATION: "mstts:audioduration"
 };
 var SSML_ATTRS = {
   VERSION: "version",
@@ -201,6 +202,9 @@ function getAttributes(element) {
     case SSML_TAGS.VISEME:
       addAttribute(attributes, SSML_ATTRS.TYPE, element.typeValue ?? element.visemeType);
       break;
+    case SSML_TAGS.MSTTS_AUDIO_DURATION:
+      addAttribute(attributes, SSML_ATTRS.VALUE, element.value);
+      break;
     case SSML_TAGS.PARAGRAPH:
     case SSML_TAGS.SENTENCE:
     case SSML_TAGS.WORD:
@@ -225,6 +229,8 @@ function getTagName(element) {
     case SSML_TAGS.VISEME:
     case SSML_TAGS.MSTTS_VISEME:
       return SSML_TAGS.MSTTS_VISEME;
+    case SSML_TAGS.MSTTS_AUDIO_DURATION:
+      return SSML_TAGS.MSTTS_AUDIO_DURATION;
     case "element":
     case "custom":
       return element.name;
@@ -757,6 +763,12 @@ function convertElement(node) {
       if (typeValue !== void 0) element.typeValue = typeValue;
       return finishElement(element, node, attributes);
     }
+    case SSML_TAGS.MSTTS_AUDIO_DURATION: {
+      const element = { type: SSML_TAGS.MSTTS_AUDIO_DURATION };
+      const value = readAttribute(attributes, SSML_ATTRS.VALUE);
+      if (value !== void 0) element.value = value;
+      return finishElement(element, node, attributes);
+    }
     default: {
       const element = {
         name: node.name,
@@ -858,6 +870,7 @@ var INTRINSICALLY_EMPTY_ELEMENTS = /* @__PURE__ */ new Set([
   "lexicon",
   "mark",
   "mstts:silence",
+  "mstts:audioduration",
   "mstts:viseme",
   "silence",
   "viseme"
@@ -1516,6 +1529,8 @@ function updateEditableText(document2, value) {
 // packages/ssml-editor-react/src/constants/ssmlPresets.ts
 var ssmlPresets_exports = {};
 __export(ssmlPresets_exports, {
+  AUDIO_DURATION_DESCRIPTIONS: () => AUDIO_DURATION_DESCRIPTIONS,
+  AUDIO_DURATION_PRESETS: () => AUDIO_DURATION_PRESETS,
   BREAK_STRENGTH_PRESETS: () => BREAK_STRENGTH_PRESETS,
   BREAK_TIME_DESCRIPTIONS: () => BREAK_TIME_DESCRIPTIONS,
   BREAK_TIME_PRESETS: () => BREAK_TIME_PRESETS,
@@ -1548,6 +1563,103 @@ __export(ssmlPresets_exports, {
   getExpressAsStyleCategory: () => getExpressAsStyleCategory,
   resolveExpressAsStyles: () => resolveExpressAsStyles
 });
+
+// packages/ssml-editor-react/src/constants/azureVoiceStyleMap.generated.ts
+var AZURE_VOICE_STYLE_MAP = {
+  "de-DE-ConradNeural": ["cheerful", "sad"],
+  "de-DE-KatjaNeural": ["cheerful", "sad"],
+  "en-US-AndrewNeural": ["empathetic", "relieved"],
+  "en-US-GuyNeural": [
+    "angry",
+    "cheerful",
+    "excited",
+    "friendly",
+    "hopeful",
+    "newscast",
+    "sad",
+    "shouting",
+    "terrified",
+    "unfriendly",
+    "whispering"
+  ],
+  "en-US-JennyMultilingualNeural": [
+    "cheerful",
+    "empathetic",
+    "excited",
+    "friendly",
+    "hopeful",
+    "sad",
+    "shouting",
+    "terrified",
+    "unfriendly",
+    "whispering"
+  ],
+  "en-US-JennyNeural": [
+    "assistant",
+    "chat",
+    "customerservice",
+    "newscast",
+    "cheerful",
+    "empathetic",
+    "excited",
+    "friendly",
+    "hopeful",
+    "sad",
+    "shouting",
+    "terrified",
+    "unfriendly",
+    "whispering"
+  ],
+  "es-ES-ElviraNeural": [],
+  "fil-PH-AngeloNeural": [],
+  "fr-FR-DeniseNeural": ["cheerful", "sad"],
+  "fr-FR-HenriNeural": ["cheerful", "sad"],
+  "id-ID-GadisNeural": [],
+  "it-IT-ElsaNeural": ["cheerful", "sad"],
+  "ja-JP-KeitaNeural": ["chat"],
+  "ja-JP-MayuNeural": ["calm", "cheerful", "sad"],
+  "ja-JP-NanamiNeural": ["chat", "customerservice", "cheerful", "whispering", "sad"],
+  "ko-KR-SunHiNeural": ["cheerful", "sad"],
+  "ms-MY-YasminNeural": [],
+  "pt-BR-FranciscaNeural": ["calm"],
+  "ru-RU-SvetlanaNeural": ["cheerful", "sad", "angry", "disgruntled", "embarrassed", "fearful"],
+  "th-TH-PremwadeeNeural": [],
+  "vi-VN-HoaiMyNeural": [],
+  "zh-CN-XiaoxiaoNeural": [
+    "assistant",
+    "chat",
+    "customerservice",
+    "newscast",
+    "cheerful",
+    "empathetic",
+    "excited",
+    "friendly",
+    "hopeful",
+    "sad",
+    "terrified",
+    "whispering",
+    "poetry-reading",
+    "sports_commentary",
+    "sports_commentary_excited",
+    "story"
+  ],
+  "zh-CN-YunxiNeural": [
+    "narration-relaxed",
+    "embarrassed",
+    "fearful",
+    "sad",
+    "disgruntled",
+    "serious",
+    "angry",
+    "depressed",
+    "chat",
+    "cheerful",
+    "assistant"
+  ],
+  "zh-TW-HsiaoChenNeural": []
+};
+
+// packages/ssml-editor-react/src/constants/ssmlPresets.ts
 var SSML_PRESETS = [
   {
     id: "basic",
@@ -1647,7 +1759,7 @@ function getExpressAsStyleCategory(style) {
   }
   return "other";
 }
-var VOICE_STYLE_MAP = {
+var LEGACY_VOICE_STYLE_MAP = {
   "ja-JP-MayuNeural": ["calm", "cheerful", "sad"],
   "ja-JP-KeitaNeural": ["chat"],
   "ja-JP-NanamiNeural": ["chat", "customerservice", "cheerful", "whispering", "sad"],
@@ -1733,6 +1845,7 @@ var VOICE_STYLE_MAP = {
   "de-DE-ConradNeural": ["cheerful", "sad"],
   "ru-RU-SvetlanaNeural": ["cheerful", "sad", "angry", "disgruntled", "embarrassed", "fearful"]
 };
+var VOICE_STYLE_MAP = Object.keys(AZURE_VOICE_STYLE_MAP).length > 0 ? AZURE_VOICE_STYLE_MAP : LEGACY_VOICE_STYLE_MAP;
 var VOICE_STYLE_MAP_BY_NORMALIZED_NAME = new Map(
   Object.entries(VOICE_STYLE_MAP).map(([voiceName, styles]) => [voiceName.toLowerCase(), styles])
 );
@@ -1775,6 +1888,7 @@ var SAY_AS_PRESETS = [
 ];
 var LANGUAGE_PRESETS = ["ja-JP", "en-US", "de-DE", "fr-FR"];
 var SILENCE_VALUE_PRESETS = ["300ms", "500ms", "1s"];
+var AUDIO_DURATION_PRESETS = ["5s", "10s", "30s"];
 var SILENCE_TYPE_PRESETS = [
   "Leading",
   "Tailing",
@@ -1783,6 +1897,11 @@ var SILENCE_TYPE_PRESETS = [
   "Semicolon",
   "Enumerationcomma"
 ];
+var AUDIO_DURATION_DESCRIPTIONS = {
+  "5s": { ja: "5\u79D2", en: "5 seconds" },
+  "10s": { ja: "10\u79D2", en: "10 seconds" },
+  "30s": { ja: "30\u79D2", en: "30 seconds" }
+};
 var PHONEME_ALPHABET_PRESETS = ["ipa", "sapi", "ups", "x-sampa"];
 var VISEME_TYPE_PRESETS = ["redlips_front", "FacialExpression"];
 var SSML_ATTRIBUTE_PRESETS = {
@@ -1826,6 +1945,9 @@ var SSML_ATTRIBUTE_PRESETS = {
     type: SILENCE_TYPE_PRESETS,
     value: SILENCE_VALUE_PRESETS
   },
+  "mstts:audioduration": {
+    value: AUDIO_DURATION_PRESETS
+  },
   silence: {
     type: SILENCE_TYPE_PRESETS,
     value: SILENCE_VALUE_PRESETS
@@ -2313,6 +2435,10 @@ var SSML_COMPLETION_SNIPPETS = [
     label: "mstts:express-as",
     insertText: `<mstts:express-as style="cheerful">\${1:text}</mstts:express-as>`
   },
+  {
+    label: "mstts:audioduration",
+    insertText: '<mstts:audioduration value="10s" />'
+  },
   {
     label: "sub",
     insertText: `<sub alias="\${1:\u8AAD\u307F}">\${2:\u6F22\u5B57}</sub>`
@@ -2580,6 +2706,13 @@ var SSML_HOVER_COPY = {
         parameters: {
           type: { title: "type", description: "\u30D3\u30BC\u30FC\u30E0\u30A4\u30D9\u30F3\u30C8\u306E\u5F62\u5F0F\u3002" }
         }
+      },
+      "mstts:audioduration": {
+        title: "\u97F3\u58F0\u9577",
+        description: "\u5408\u6210\u97F3\u58F0\u306E\u76EE\u6A19\u6642\u9593\u3092\u6307\u5B9A\u3057\u307E\u3059\u3002",
+        parameters: {
+          value: { title: "value", description: "\u76EE\u6A19\u6642\u9593\u3002\u4F8B: `10s`\u3001`5000ms`\u3001`00:00:10`\u3002" }
+        }
       }
     }
   },
@@ -2737,6 +2870,13 @@ var SSML_HOVER_COPY = {
         parameters: {
           type: { title: "type", description: "The viseme event format." }
         }
+      },
+      "mstts:audioduration": {
+        title: "Audio duration",
+        description: "Sets the target duration of synthesized audio.",
+        parameters: {
+          value: { title: "value", description: "The target duration, such as `10s`, `5000ms`, or `00:00:10`." }
+        }
       }
     }
   }
@@ -3033,6 +3173,17 @@ var SSML_TAG_DEFINITIONS = [
         values: VISEME_TYPE_PRESETS2
       }
     ]
+  },
+  {
+    name: "mstts:audioduration",
+    description: "Sets the target duration of synthesized audio.",
+    parameters: [
+      {
+        name: "value",
+        description: "The target duration, such as `10s`, `5000ms`, or `00:00:10`.",
+        example: "10s"
+      }
+    ]
   }
 ];
 var definitionsByName = /* @__PURE__ */ new Map();
@@ -3574,13 +3725,34 @@ var SSML_INSERTIONS = [
       suffix: "",
       mode: "insert"
     })
+  },
+  {
+    id: "mstts:audioduration",
+    icon: "\u25F7",
+    tagName: "mstts:audioduration",
+    selfClosing: true,
+    labels: { ja: "\u97F3\u58F0\u9577", en: "Audio duration" },
+    descriptions: {
+      ja: "\u5408\u6210\u97F3\u58F0\u306E\u76EE\u6A19\u6642\u9593\u3092\u6307\u5B9A\u3057\u307E\u3059\u3002",
+      en: "Sets the target duration of synthesized audio."
+    },
+    parameterDescription: {
+      ja: "\u76EE\u6A19\u6642\u9593\u3092\u9078\u629E\u3057\u307E\u3059\u3002",
+      en: "Selects the target duration."
+    },
+    options: createInsertionOptions(AUDIO_DURATION_PRESETS, AUDIO_DURATION_DESCRIPTIONS),
+    createTemplate: (value) => ({
+      prefix: `<mstts:audioduration value="${value}"/>`,
+      suffix: "",
+      mode: "insert"
+    })
   }
 ];
 var DEFAULT_INSERTION_GROUPS = [
   {
     id: "pauses",
     labels: { ja: "\u9593\u30FB\u7121\u97F3", en: "Pauses" },
-    insertionIds: ["break", "mstts:silence"]
+    insertionIds: ["break", "mstts:silence", "mstts:audioduration"]
   },
   {
     id: "prosody",