ssml-builder-js 2.8.0 → 2.8.1

package/dist/index.js

@@ -44,6 +44,7 @@ __export(src_exports, {
   buildPartialSsml: () => buildPartialSsml,
   buildSsml: () => buildSsml,
   extractSsmlText: () => extractSsmlText,
+  isValidAzureAudioDuration: () => isValidAzureAudioDuration,
   mapSsmlTextNodes: () => mapSsmlTextNodes,
   normalizeAzureLanguage: () => normalizeAzureLanguage,
   parseSsml: () => parseSsml,
@@ -85,7 +86,8 @@ 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",
@@ -214,6 +216,9 @@ 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:
@@ -238,6 +243,8 @@ 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;
@@ -772,6 +779,12 @@ 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,
@@ -1046,93 +1059,126 @@ async function mapSsmlTextNodes(ssml, transform, options = {}) {
   return result + ssml.slice(cursor);
 }
 
+// packages/ssml-core/src/generated/azureVoiceDefinitions.ts
+var AZURE_VOICE_DEFINITIONS = [
+  { name: "de-DE-ConradNeural", locale: "de-DE", styles: ["cheerful", "sad"] },
+  { name: "de-DE-KatjaNeural", locale: "de-DE", styles: ["cheerful", "sad"] },
+  { name: "en-US-AndrewNeural", locale: "en-US", styles: ["empathetic", "relieved"] },
+  {
+    name: "en-US-GuyNeural",
+    locale: "en-US",
+    styles: [
+      "angry",
+      "cheerful",
+      "excited",
+      "friendly",
+      "hopeful",
+      "newscast",
+      "sad",
+      "shouting",
+      "terrified",
+      "unfriendly",
+      "whispering"
+    ]
+  },
+  {
+    name: "en-US-JennyMultilingualNeural",
+    locale: "en-US",
+    styles: [
+      "cheerful",
+      "empathetic",
+      "excited",
+      "friendly",
+      "hopeful",
+      "sad",
+      "shouting",
+      "terrified",
+      "unfriendly",
+      "whispering"
+    ]
+  },
+  {
+    name: "en-US-JennyNeural",
+    locale: "en-US",
+    styles: [
+      "assistant",
+      "chat",
+      "customerservice",
+      "newscast",
+      "cheerful",
+      "empathetic",
+      "excited",
+      "friendly",
+      "hopeful",
+      "sad",
+      "shouting",
+      "terrified",
+      "unfriendly",
+      "whispering"
+    ]
+  },
+  { name: "es-ES-ElviraNeural", locale: "es-ES" },
+  { name: "fil-PH-AngeloNeural", locale: "fil-PH" },
+  { name: "fr-FR-DeniseNeural", locale: "fr-FR", styles: ["cheerful", "sad"] },
+  { name: "fr-FR-HenriNeural", locale: "fr-FR", styles: ["cheerful", "sad"] },
+  { name: "id-ID-GadisNeural", locale: "id-ID" },
+  { name: "it-IT-ElsaNeural", locale: "it-IT", styles: ["cheerful", "sad"] },
+  { name: "ja-JP-KeitaNeural", locale: "ja-JP", styles: ["chat"] },
+  { name: "ja-JP-MayuNeural", locale: "ja-JP", styles: ["calm", "cheerful", "sad"] },
+  { name: "ja-JP-NanamiNeural", locale: "ja-JP", styles: ["chat", "customerservice", "cheerful", "whispering", "sad"] },
+  { name: "ko-KR-SunHiNeural", locale: "ko-KR", styles: ["cheerful", "sad"] },
+  { name: "ms-MY-YasminNeural", locale: "ms-MY" },
+  { name: "pt-BR-FranciscaNeural", locale: "pt-BR", styles: ["calm"] },
+  {
+    name: "ru-RU-SvetlanaNeural",
+    locale: "ru-RU",
+    styles: ["cheerful", "sad", "angry", "disgruntled", "embarrassed", "fearful"]
+  },
+  { name: "th-TH-PremwadeeNeural", locale: "th-TH" },
+  { name: "vi-VN-HoaiMyNeural", locale: "vi-VN" },
+  {
+    name: "zh-CN-XiaoxiaoNeural",
+    locale: "zh-CN",
+    styles: [
+      "assistant",
+      "chat",
+      "customerservice",
+      "newscast",
+      "cheerful",
+      "empathetic",
+      "excited",
+      "friendly",
+      "hopeful",
+      "sad",
+      "terrified",
+      "whispering",
+      "poetry-reading",
+      "sports_commentary",
+      "sports_commentary_excited",
+      "story"
+    ]
+  },
+  {
+    name: "zh-CN-YunxiNeural",
+    locale: "zh-CN",
+    styles: [
+      "narration-relaxed",
+      "embarrassed",
+      "fearful",
+      "sad",
+      "disgruntled",
+      "serious",
+      "angry",
+      "depressed",
+      "chat",
+      "cheerful",
+      "assistant"
+    ]
+  },
+  { name: "zh-TW-HsiaoChenNeural", locale: "zh-TW" }
+];
+
 // packages/ssml-core/src/azureValidation.ts
-var EXPRESS_AS_STYLES = {
-  "en-us-jennyneural": [
-    "assistant",
-    "chat",
-    "customerservice",
-    "newscast",
-    "cheerful",
-    "empathetic",
-    "excited",
-    "friendly",
-    "hopeful",
-    "sad",
-    "shouting",
-    "terrified",
-    "unfriendly",
-    "whispering"
-  ],
-  "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-andrewneural": ["empathetic", "relieved"],
-  "ja-jp-mayuneural": ["calm", "cheerful", "sad"],
-  "ja-jp-nanamineural": ["chat", "customerservice", "cheerful", "whispering", "sad"],
-  "ja-jp-keitaneural": ["chat"],
-  "ko-kr-sunhineural": ["cheerful", "sad"],
-  "zh-cn-yunxineural": [
-    "narration-relaxed",
-    "embarrassed",
-    "fearful",
-    "sad",
-    "disgruntled",
-    "serious",
-    "angry",
-    "depressed",
-    "chat",
-    "cheerful",
-    "assistant"
-  ],
-  "zh-cn-xiaoxiaoneural": [
-    "assistant",
-    "chat",
-    "customerservice",
-    "newscast",
-    "cheerful",
-    "empathetic",
-    "excited",
-    "friendly",
-    "hopeful",
-    "sad",
-    "terrified",
-    "whispering",
-    "poetry-reading",
-    "sports_commentary",
-    "sports_commentary_excited",
-    "story"
-  ],
-  "fr-fr-deniseneural": ["cheerful", "sad"],
-  "fr-fr-henrineural": ["cheerful", "sad"],
-  "pt-br-franciscaneural": ["calm"],
-  "it-it-elsaneural": ["cheerful", "sad"],
-  "de-de-katjaneural": ["cheerful", "sad"],
-  "de-de-conradneural": ["cheerful", "sad"],
-  "ru-ru-svetlananeural": ["cheerful", "sad", "angry", "disgruntled", "embarrassed", "fearful"]
-};
 var ALLOWED_BREAK_STRENGTHS = /* @__PURE__ */ new Set(["none", "x-weak", "weak", "medium", "strong", "x-strong"]);
 var ALLOWED_SAY_AS = /* @__PURE__ */ new Set([
   "characters",
@@ -1261,18 +1307,17 @@ function isSupportedProsodyRate(value) {
   const numericValue = Number(multiplier[1]);
   return numericValue >= 0.5 && numericValue <= 2;
 }
+function isValidAzureAudioDuration(value) {
+  const trimmed = value.trim();
+  const numeric = /^(\d+(?:\.\d+)?)(ms|s)$/.exec(trimmed);
+  if (numeric) return Number(numeric[1]) > 0;
+  const clock = /^(\d{2,}):([0-5]\d):([0-5]\d)(?:\.(\d{1,3}))?$/.exec(trimmed);
+  if (!clock) return false;
+  return Number(clock[1]) > 0 || Number(clock[2]) > 0 || Number(clock[3]) > 0 || Number(clock[4] ?? 0) > 0;
+}
 function attr(token, name) {
   return token.attributes.get(name.toLowerCase());
 }
-var ADDITIONAL_VOICE_DEFINITIONS = [
-  { name: "zh-TW-HsiaoChenNeural", locale: "zh-TW" },
-  { name: "es-ES-ElviraNeural", locale: "es-ES" },
-  { name: "th-TH-PremwadeeNeural", locale: "th-TH" },
-  { name: "fil-PH-AngeloNeural", locale: "fil-PH" },
-  { name: "vi-VN-HoaiMyNeural", locale: "vi-VN" },
-  { name: "id-ID-GadisNeural", locale: "id-ID" },
-  { name: "ms-MY-YasminNeural", locale: "ms-MY" }
-];
 var DEFAULT_LANGUAGE_ALIASES = {
   "zh-CN": ["zh-Hans"],
   "zh-TW": ["zh-Hant"]
@@ -1335,10 +1380,7 @@ function definitionFromStyleMap(voiceName, styles) {
 }
 function normalizeVoiceCatalog(options) {
   const definitions = /* @__PURE__ */ new Map();
-  for (const [name, styles] of Object.entries(EXPRESS_AS_STYLES)) {
-    definitions.set(name.toLowerCase(), definitionFromStyleMap(name, styles));
-  }
-  for (const definition of ADDITIONAL_VOICE_DEFINITIONS) definitions.set(definition.name.toLowerCase(), definition);
+  for (const definition of AZURE_VOICE_DEFINITIONS) definitions.set(definition.name.toLowerCase(), definition);
   for (const definition of options.voiceCatalog ?? []) definitions.set(definition.name.toLowerCase(), definition);
   for (const definition of options.voiceDefinitions ?? []) definitions.set(definition.name.toLowerCase(), definition);
   for (const definition of options.customVoiceDefinitions ?? [])
@@ -1475,6 +1517,18 @@ function validateElement(token, source, diagnostics, voiceName, options, voiceCa
     if (!value || !/^\d+(?:\.\d+)?(?:ms|s)$/.test(value.trim()))
       addDiagnostic(diagnostics, source, token.start, '<mstts:silence> requires a time-valued "value" attribute.');
   }
+  if (name === "mstts:audioduration") {
+    const value = attr(token, "value");
+    if (!value || !isValidAzureAudioDuration(value))
+      addDiagnostic(
+        diagnostics,
+        source,
+        token.start,
+        '<mstts:audioduration> requires a positive duration such as "10s", "5000ms", or "00:00:10".'
+      );
+    if (!token.selfClosing)
+      addDiagnostic(diagnostics, source, token.start, "<mstts:audioduration> must be self-closing.");
+  }
   if (name === "mstts:viseme") {
     const type = attr(token, "type");
     if (!type || !ALLOWED_VISEME_TYPES.has(type))
@@ -1765,6 +1819,7 @@ _options = new WeakMap();
   buildPartialSsml,
   buildSsml,
   extractSsmlText,
+  isValidAzureAudioDuration,
   mapSsmlTextNodes,
   normalizeAzureLanguage,
   parseSsml,