ssml-builder-js 2.8.1 → 2.9.0

package/dist/{chunk-MWSDFGXW.mjs → chunk-JNJVTEL6.mjs}

@@ -959,6 +959,11 @@ var SSML_ATTRIBUTE_PRESETS = {
   "mstts:audioduration": {
     value: AUDIO_DURATION_PRESETS
   },
+  "mstts:backgroundaudio": {
+    volume: PROSODY_VOLUME_PRESETS,
+    fadein: AUDIO_DURATION_PRESETS,
+    fadeout: AUDIO_DURATION_PRESETS
+  },
   silence: {
     type: SILENCE_TYPE_PRESETS,
     value: SILENCE_VALUE_PRESETS
@@ -1343,7 +1348,13 @@ var SSML_TAGS = {
   SILENCE: "silence",
   MSTTS_VISEME: "mstts:viseme",
   VISEME: "viseme",
-  MSTTS_AUDIO_DURATION: "mstts:audioduration"
+  MSTTS_AUDIO_DURATION: "mstts:audioduration",
+  MSTTS_DIALOG: "mstts:dialog",
+  MSTTS_TURN: "mstts:turn",
+  MSTTS_BACKGROUND_AUDIO: "mstts:backgroundaudio",
+  MSTTS_TTS_EMBEDDING: "mstts:ttsembedding",
+  MSTTS_EMBEDDING: "mstts:embedding",
+  MSTTS_VOICE_CONVERSION: "mstts:voiceconversion"
 };
 var SSML_ATTRS = {
   VERSION: "version",
@@ -1352,6 +1363,7 @@ var SSML_ATTRS = {
   LANG: "lang",
   MSTTS_XMLNS: "xmlns:mstts",
   NAME: "name",
+  VOICE: "voice",
   EFFECT: "effect",
   RATE: "rate",
   PITCH: "pitch",
@@ -1383,7 +1395,9 @@ var SSML_ATTRS = {
   MARK: "mark",
   URI: "uri",
   TYPE: "type",
-  VALUE: "value"
+  VALUE: "value",
+  FADE_IN: "fadein",
+  FADE_OUT: "fadeout"
 };
 function escapeText(value) {
   return value.replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;");
@@ -1473,6 +1487,20 @@ function getAttributes(element) {
     case SSML_TAGS.MSTTS_AUDIO_DURATION:
       addAttribute(attributes, SSML_ATTRS.VALUE, element.value);
       break;
+    case SSML_TAGS.MSTTS_TURN:
+      addAttribute(attributes, SSML_ATTRS.VOICE, element.voice);
+      break;
+    case SSML_TAGS.MSTTS_BACKGROUND_AUDIO:
+      addAttribute(attributes, SSML_ATTRS.SRC, element.src);
+      addAttribute(attributes, SSML_ATTRS.VOLUME, element.volume);
+      addAttribute(attributes, SSML_ATTRS.FADE_IN, element.fadeIn ?? element.fadein);
+      addAttribute(attributes, SSML_ATTRS.FADE_OUT, element.fadeOut ?? element.fadeout);
+      break;
+    case SSML_TAGS.MSTTS_DIALOG:
+    case SSML_TAGS.MSTTS_TTS_EMBEDDING:
+    case SSML_TAGS.MSTTS_EMBEDDING:
+    case SSML_TAGS.MSTTS_VOICE_CONVERSION:
+      break;
     case SSML_TAGS.PARAGRAPH:
     case SSML_TAGS.SENTENCE:
     case SSML_TAGS.WORD:
@@ -2037,6 +2065,40 @@ function convertElement(node) {
       if (value !== void 0) element.value = value;
       return finishElement(element, node, attributes);
     }
+    case SSML_TAGS.MSTTS_DIALOG: {
+      const element = { type: SSML_TAGS.MSTTS_DIALOG };
+      return finishElement(element, node, attributes);
+    }
+    case SSML_TAGS.MSTTS_TURN: {
+      const element = { type: SSML_TAGS.MSTTS_TURN };
+      const voice = readAttribute(attributes, SSML_ATTRS.VOICE);
+      if (voice !== void 0) element.voice = voice;
+      return finishElement(element, node, attributes);
+    }
+    case SSML_TAGS.MSTTS_BACKGROUND_AUDIO: {
+      const element = { type: SSML_TAGS.MSTTS_BACKGROUND_AUDIO };
+      const src = readAttribute(attributes, SSML_ATTRS.SRC);
+      const volume = readAttribute(attributes, SSML_ATTRS.VOLUME);
+      const fadeIn = readAttribute(attributes, SSML_ATTRS.FADE_IN);
+      const fadeOut = readAttribute(attributes, SSML_ATTRS.FADE_OUT);
+      if (src !== void 0) element.src = src;
+      if (volume !== void 0) element.volume = volume;
+      if (fadeIn !== void 0) element.fadeIn = fadeIn;
+      if (fadeOut !== void 0) element.fadeOut = fadeOut;
+      return finishElement(element, node, attributes);
+    }
+    case SSML_TAGS.MSTTS_TTS_EMBEDDING: {
+      const element = { type: SSML_TAGS.MSTTS_TTS_EMBEDDING };
+      return finishElement(element, node, attributes);
+    }
+    case SSML_TAGS.MSTTS_EMBEDDING: {
+      const element = { type: SSML_TAGS.MSTTS_EMBEDDING };
+      return finishElement(element, node, attributes);
+    }
+    case SSML_TAGS.MSTTS_VOICE_CONVERSION: {
+      const element = { type: SSML_TAGS.MSTTS_VOICE_CONVERSION };
+      return finishElement(element, node, attributes);
+    }
     default: {
       const element = {
         name: node.name,
@@ -2157,6 +2219,644 @@ function validateSsml(xmlString) {
     };
   }
 }
+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" }
+];
+var ALLOWED_BREAK_STRENGTHS = /* @__PURE__ */ new Set(["none", "x-weak", "weak", "medium", "strong", "x-strong"]);
+var ALLOWED_SAY_AS = /* @__PURE__ */ new Set([
+  "characters",
+  "spell-out",
+  "cardinal",
+  "ordinal",
+  "number",
+  "date",
+  "time",
+  "telephone",
+  "fraction",
+  "address",
+  "name",
+  "currency",
+  "number_digit"
+]);
+var ALLOWED_ROLES = /* @__PURE__ */ new Set([
+  "Girl",
+  "Boy",
+  "YoungAdultFemale",
+  "YoungAdultMale",
+  "OlderAdultFemale",
+  "OlderAdultMale",
+  "SeniorFemale",
+  "SeniorMale"
+]);
+var ALLOWED_EMPHASIS_LEVELS = /* @__PURE__ */ new Set(["strong", "moderate", "reduced", "none"]);
+var ALLOWED_SILENCE_TYPES = /* @__PURE__ */ new Set([
+  "Leading",
+  "Tailing",
+  "Sentenceboundary",
+  "Comma",
+  "Semicolon",
+  "Enumerationcomma"
+]);
+var ALLOWED_VISEME_TYPES = /* @__PURE__ */ new Set(["redlips_front", "FacialExpression"]);
+function decodeAttribute(value) {
+  return value.replace(
+    /&(?:amp|apos|gt|lt|quot);/gi,
+    (entity) => ({ "&amp;": "&", "&apos;": "'", "&gt;": ">", "&lt;": "<", "&quot;": '"' })[entity.toLowerCase()] ?? entity
+  );
+}
+function findTagEnd2(source, start) {
+  let quote = "";
+  for (let index = start; index < source.length; index += 1) {
+    const character = source[index];
+    if (quote) {
+      if (character === quote) quote = "";
+    } else if (character === '"' || character === "'") quote = character;
+    else if (character === ">") return index;
+  }
+  return source.length - 1;
+}
+function tokenizeElements(source) {
+  const tokens = [];
+  const openElements = [];
+  let index = 0;
+  while (index < source.length) {
+    const start = source.indexOf("<", index);
+    if (start === -1) break;
+    if (source.startsWith("<!--", start)) {
+      const end2 = source.indexOf("-->", start + 4);
+      index = end2 === -1 ? source.length : end2 + 3;
+      continue;
+    }
+    if (source.startsWith("<![CDATA[", start)) {
+      const end2 = source.indexOf("]]>", start + 9);
+      index = end2 === -1 ? source.length : end2 + 3;
+      continue;
+    }
+    if (source.startsWith("<?", start)) {
+      const end2 = source.indexOf("?>", start + 2);
+      index = end2 === -1 ? source.length : end2 + 2;
+      continue;
+    }
+    const end = findTagEnd2(source, start + 1);
+    const raw = source.slice(start, end + 1);
+    if (raw.startsWith("</")) {
+      openElements.pop();
+      index = end + 1;
+      continue;
+    }
+    const nameMatch = /^<\s*([A-Za-z_][A-Za-z0-9_.:-]*)/.exec(raw);
+    if (!nameMatch?.[1]) {
+      index = end + 1;
+      continue;
+    }
+    const attributes = /* @__PURE__ */ new Map();
+    const attributeSource = raw.slice(nameMatch[0].length, raw.length - 1).replace(/\/\s*$/, "");
+    const attributePattern = /([A-Za-z_][A-Za-z0-9_.:-]*)\s*=\s*(["'])([\s\S]*?)\2/g;
+    for (const match of attributeSource.matchAll(attributePattern)) {
+      attributes.set(match[1].toLowerCase(), decodeAttribute(match[3]));
+    }
+    const selfClosing = /\/\s*>$/.test(raw);
+    const parent = openElements[openElements.length - 1];
+    const parentVoiceName = [...openElements].reverse().find((element) => element.voiceName)?.voiceName;
+    const tokenName = nameMatch[1];
+    const tokenVoiceName = tokenName.toLowerCase() === "voice" ? attributes.get("name") : tokenName.toLowerCase() === "mstts:turn" ? attributes.get("voice") ?? parentVoiceName : parentVoiceName;
+    tokens.push({
+      attributes,
+      end,
+      name: tokenName,
+      parentName: parent?.name,
+      parentVoiceName,
+      selfClosing,
+      start
+    });
+    if (!selfClosing) {
+      openElements.push({
+        name: tokenName,
+        voiceName: tokenVoiceName
+      });
+    }
+    index = end + 1;
+  }
+  return tokens;
+}
+function location(source, offset) {
+  const before = source.slice(0, Math.max(0, offset));
+  const line = before.split("\n").length;
+  return { line, column: before.length - (before.lastIndexOf("\n") + 1) + 1 };
+}
+function addDiagnostic(diagnostics, source, offset, message, severity = "error", code2) {
+  diagnostics.push({
+    ...location(source, offset),
+    message,
+    severity,
+    source: "ssml-static-validator",
+    ...code2 ? { code: code2 } : {}
+  });
+}
+function isSupportedProsodyRate(value) {
+  const trimmed = value.trim();
+  if (/^(x-slow|slow|medium|fast|x-fast|[+-]?\d+(?:\.\d+)?%)$/.test(trimmed)) return true;
+  const multiplier = /^(\d+(?:\.\d+)?)(x)?$/i.exec(trimmed);
+  if (!multiplier) return false;
+  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 DEFAULT_LANGUAGE_ALIASES = {
+  "zh-CN": ["zh-Hans"],
+  "zh-TW": ["zh-Hant"]
+};
+function canonicalLanguageTag(language) {
+  const trimmed = language.trim();
+  if (!trimmed) return "";
+  try {
+    return new Intl.Locale(trimmed).toString().toLowerCase();
+  } catch {
+    return trimmed.toLowerCase();
+  }
+}
+function createLanguageNormalizer(options) {
+  const aliases = /* @__PURE__ */ new Map();
+  const addAliasGroup = (canonical, values) => {
+    const normalizedCanonical = canonicalLanguageTag(canonical);
+    if (!normalizedCanonical) return;
+    aliases.set(normalizedCanonical, normalizedCanonical);
+    for (const value of values) {
+      const normalizedValue = canonicalLanguageTag(value);
+      if (normalizedValue) aliases.set(normalizedValue, normalizedCanonical);
+    }
+  };
+  for (const [canonical, values] of Object.entries(DEFAULT_LANGUAGE_ALIASES)) addAliasGroup(canonical, values);
+  for (const [canonical, valueOrValues] of Object.entries(options.languageAliases ?? {}))
+    addAliasGroup(canonical, typeof valueOrValues === "string" ? [valueOrValues] : valueOrValues);
+  return (language) => {
+    const customValue = options.normalizeLanguage ? options.normalizeLanguage(language) : language;
+    const normalized = canonicalLanguageTag(customValue);
+    return aliases.get(normalized) ?? normalized;
+  };
+}
+function voiceLocalePrefix(voiceName) {
+  const match = /^(?<language>[A-Za-z]{2,3})-(?<region>[A-Za-z]{2}|\d{3})(?:-|$)/.exec(voiceName.trim());
+  if (!match?.groups) return void 0;
+  const tag = `${match.groups.language}-${match.groups.region}`;
+  return {
+    language: match.groups.language.toLowerCase(),
+    region: match.groups.region.toLowerCase(),
+    tag
+  };
+}
+function definitionFromStyleMap(voiceName, styles) {
+  return {
+    name: voiceName,
+    locale: voiceLocalePrefix(voiceName)?.tag ?? "",
+    styles
+  };
+}
+function normalizeVoiceCatalog(options) {
+  const definitions = /* @__PURE__ */ new Map();
+  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 ?? [])
+    definitions.set(definition.name.toLowerCase(), definition);
+  for (const [voiceName, styles] of Object.entries(options.customVoiceStyleMap ?? {})) {
+    const key = voiceName.toLowerCase();
+    const current = definitions.get(key);
+    definitions.set(key, {
+      ...current ?? definitionFromStyleMap(voiceName, styles),
+      name: current?.name ?? voiceName,
+      styles: styles.map((style) => style.toLowerCase())
+    });
+  }
+  return definitions;
+}
+function diagnosticSeverity(policy) {
+  if (policy === "ignore") return void 0;
+  return policy === "error" ? "error" : "warning";
+}
+function languagePart(language) {
+  try {
+    return new Intl.Locale(language).language.toLowerCase();
+  } catch {
+    return language.split("-")[0]?.toLowerCase() ?? "";
+  }
+}
+function definitionMatchesLanguage(definition, voiceName, language, normalizeLanguage) {
+  const candidateLanguages = definition ? [definition.locale, ...definition.secondaryLocales ?? []].filter(Boolean) : [voiceLocalePrefix(voiceName)?.tag ?? ""];
+  if (candidateLanguages.length === 0 || !language.trim()) return void 0;
+  const normalizedLanguage = normalizeLanguage(language);
+  const normalizedCandidates = candidateLanguages.map(normalizeLanguage);
+  if (normalizedCandidates.includes(normalizedLanguage)) return true;
+  if (!normalizedLanguage || !normalizedCandidates.some(Boolean)) return void 0;
+  return normalizedLanguage === languagePart(normalizedLanguage) ? normalizedCandidates.some((candidate) => languagePart(candidate) === normalizedLanguage) : false;
+}
+function canonicalTagName(name) {
+  const normalized = name.toLowerCase();
+  if (normalized === "express-as" || normalized === "expressas") return "mstts:express-as";
+  if (normalized === "sayas") return "say-as";
+  return normalized;
+}
+function validateVoiceFeatureMatrix(token, source, diagnostics, voiceName, definition) {
+  if (!voiceName || !definition || token.name.toLowerCase() === "voice" || token.name.toLowerCase() === "mstts:turn")
+    return;
+  const tagName = canonicalTagName(token.name);
+  const unsupportedTags = new Set((definition.unsupportedTags ?? []).map(canonicalTagName));
+  const supportedTags = definition.supportedTags?.map(canonicalTagName);
+  if (unsupportedTags.has(tagName) || supportedTags !== void 0 && !supportedTags.includes(tagName)) {
+    addDiagnostic(
+      diagnostics,
+      source,
+      token.start,
+      `Tag <${token.name}> is not supported by voice "${voiceName}" according to the configured feature matrix.`,
+      "error",
+      "azure-unsupported-tag-for-voice"
+    );
+  }
+}
+function validateAudioSource(token, source, diagnostics, options, elementName2) {
+  const src = attr(token, "src");
+  if (!src) {
+    addDiagnostic(diagnostics, source, token.start, `<${elementName2}> requires a "src" attribute.`);
+    return;
+  }
+  let parsed;
+  try {
+    parsed = new URL(src);
+  } catch {
+    addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> must be an absolute HTTP(S) URL.`);
+    return;
+  }
+  if (parsed.protocol !== "https:" && !(options.allowHttpAudio && parsed.protocol === "http:"))
+    addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> must use HTTPS.`);
+  if (options.allowedAudioOrigins && !options.allowedAudioOrigins.includes(parsed.origin))
+    addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> origin "${parsed.origin}" is not allowed.`);
+  else if (!options.allowExternalAudio)
+    addDiagnostic(
+      diagnostics,
+      source,
+      token.start,
+      `<${elementName2} src> external origin "${parsed.origin}" is blocked by default; set allowExternalAudio to true or provide allowedAudioOrigins.`
+    );
+}
+function validateElement(token, source, diagnostics, voiceName, options, voiceCatalog) {
+  const name = token.name.toLowerCase();
+  if (name === "voice" && !attr(token, "name")?.trim())
+    addDiagnostic(diagnostics, source, token.start, '<voice> requires a non-empty "name" attribute.');
+  if (name === "break") {
+    const time = attr(token, "time");
+    const strength = attr(token, "strength");
+    if (!time && !strength)
+      addDiagnostic(diagnostics, source, token.start, '<break> requires either "time" or "strength".');
+    if (time && strength)
+      addDiagnostic(diagnostics, source, token.start, '<break> must not specify both "time" and "strength".');
+    if (time && !/^\d+(?:\.\d+)?(?:ms|s)$/.test(time.trim()))
+      addDiagnostic(diagnostics, source, token.start, '<break time> must use a numeric value followed by "ms" or "s".');
+    if (strength && !ALLOWED_BREAK_STRENGTHS.has(strength))
+      addDiagnostic(diagnostics, source, token.start, `Unsupported <break strength> value "${strength}".`);
+  }
+  if (name === "prosody") {
+    const rate = attr(token, "rate");
+    const pitch = attr(token, "pitch");
+    const volume = attr(token, "volume");
+    if (rate && !isSupportedProsodyRate(rate))
+      addDiagnostic(diagnostics, source, token.start, `Unsupported <prosody rate> value "${rate}".`);
+    if (pitch && !/^(x-low|low|medium|high|x-high|[+-]?\d+(?:\.\d+)?(?:st|Hz|%)?)$/.test(pitch.trim()))
+      addDiagnostic(diagnostics, source, token.start, `Unsupported <prosody pitch> value "${pitch}".`);
+    if (volume && !/^(silent|x-soft|soft|medium|loud|x-loud|[+-]?\d+(?:\.\d+)?(?:dB|%)?)$/.test(volume.trim()))
+      addDiagnostic(diagnostics, source, token.start, `Unsupported <prosody volume> value "${volume}".`);
+  }
+  if (name === "mstts:express-as" || name === "express-as" || name === "expressas") {
+    const style = attr(token, "style");
+    if (!style?.trim())
+      addDiagnostic(diagnostics, source, token.start, '<mstts:express-as> requires a non-empty "style" attribute.');
+    const degree = attr(token, "styledegree") ?? attr(token, "style-degree");
+    if (degree && (!/^\d+(?:\.\d+)?$/.test(degree) || Number(degree) < 0.01 || Number(degree) > 2))
+      addDiagnostic(
+        diagnostics,
+        source,
+        token.start,
+        "<mstts:express-as styledegree> must be a number between 0.01 and 2."
+      );
+    const role = attr(token, "role");
+    if (role && !ALLOWED_ROLES.has(role))
+      addDiagnostic(diagnostics, source, token.start, `Unsupported <mstts:express-as role> value "${role}".`);
+    const definition = voiceName ? voiceCatalog.get(voiceName.toLowerCase()) : void 0;
+    const supportedStyles = definition?.styles;
+    const severity = diagnosticSeverity(options.unsupportedStylePolicy ?? options.unknownVoicePolicy ?? "warn");
+    if (style && definition && !supportedStyles?.some((candidate) => candidate.toLowerCase() === style.toLowerCase()) && severity)
+      addDiagnostic(
+        diagnostics,
+        source,
+        token.start,
+        `Unknown style "${style}" is not supported by voice "${voiceName}" according to the configured voice style map.`,
+        severity,
+        "azure-unsupported-style"
+      );
+    if (style && voiceName && !definition && severity)
+      addDiagnostic(
+        diagnostics,
+        source,
+        token.start,
+        `Unknown style "${style}" cannot be verified because voice "${voiceName}" is not registered in the voice style map.`,
+        severity
+      );
+  }
+  if (name === "say-as" || name === "sayas") {
+    const interpretAs = attr(token, "interpret-as");
+    if (!interpretAs || !ALLOWED_SAY_AS.has(interpretAs))
+      addDiagnostic(diagnostics, source, token.start, `<say-as> requires a supported "interpret-as" value.`);
+  }
+  if (name === "phoneme" && (!attr(token, "alphabet") || !attr(token, "ph")))
+    addDiagnostic(diagnostics, source, token.start, '<phoneme> requires both "alphabet" and "ph" attributes.');
+  if (name === "emphasis" && attr(token, "level") && !ALLOWED_EMPHASIS_LEVELS.has(attr(token, "level") ?? ""))
+    addDiagnostic(diagnostics, source, token.start, `Unsupported <emphasis level> value "${attr(token, "level")}".`);
+  if (name === "sub" && !attr(token, "alias")?.trim())
+    addDiagnostic(diagnostics, source, token.start, '<sub> requires a non-empty "alias" attribute.');
+  if (name === "lang" && !attr(token, "xml:lang")?.trim() && !attr(token, "lang")?.trim())
+    addDiagnostic(diagnostics, source, token.start, '<lang> requires an "xml:lang" attribute.');
+  if (name === "mark" && !attr(token, "name")?.trim())
+    addDiagnostic(diagnostics, source, token.start, '<mark> requires a non-empty "name" attribute.');
+  if (name === "bookmark" && !attr(token, "mark")?.trim())
+    addDiagnostic(diagnostics, source, token.start, '<bookmark> requires a non-empty "mark" attribute.');
+  if (name === "lexicon") {
+    const uri = attr(token, "uri");
+    if (!uri) addDiagnostic(diagnostics, source, token.start, '<lexicon> requires a "uri" attribute.');
+    else {
+      try {
+        const parsed = new URL(uri);
+        if (parsed.protocol !== "https:")
+          addDiagnostic(diagnostics, source, token.start, "<lexicon uri> must use HTTPS.");
+      } catch {
+        addDiagnostic(diagnostics, source, token.start, "<lexicon uri> must be an absolute HTTPS URL.");
+      }
+    }
+  }
+  if (name === "mstts:silence") {
+    const type = attr(token, "type");
+    const value = attr(token, "value");
+    if (!type || !ALLOWED_SILENCE_TYPES.has(type))
+      addDiagnostic(diagnostics, source, token.start, '<mstts:silence> requires a supported "type" attribute.');
+    if (!value || !/^\d+(?:\.\d+)?(?:ms|s)$/.test(value.trim()))
+      addDiagnostic(diagnostics, source, token.start, '<mstts:silence> requires a time-valued "value" attribute.');
+  }
+  if (name === "mstts:audioduration") {
+    const value = attr(token, "value");
+    if (!value || !isValidAzureAudioDuration(value))
+      addDiagnostic(
+        diagnostics,
+        source,
+        token.start,
+        '<mstts:audioduration> requires a positive duration such as "10s", "5000ms", or "00:00:10".'
+      );
+    if (!token.selfClosing)
+      addDiagnostic(diagnostics, source, token.start, "<mstts:audioduration> must be self-closing.");
+  }
+  if (name === "mstts:viseme") {
+    const type = attr(token, "type");
+    if (!type || !ALLOWED_VISEME_TYPES.has(type))
+      addDiagnostic(diagnostics, source, token.start, '<mstts:viseme> requires a supported "type" attribute.');
+  }
+  if (name === "audio") {
+    validateAudioSource(token, source, diagnostics, options, "audio");
+  }
+  if (name === "mstts:turn") {
+    if (!attr(token, "voice")?.trim())
+      addDiagnostic(diagnostics, source, token.start, '<mstts:turn> requires a non-empty "voice" attribute.');
+    if (token.parentName?.toLowerCase() !== "mstts:dialog")
+      addDiagnostic(diagnostics, source, token.start, "<mstts:turn> is only allowed directly inside <mstts:dialog>.");
+  }
+  if (name === "mstts:backgroundaudio") {
+    validateAudioSource(token, source, diagnostics, options, "mstts:backgroundaudio");
+    const volume = attr(token, "volume");
+    if (volume && !/^(silent|x-soft|soft|medium|loud|x-loud|[+-]?\d+(?:\.\d+)?(?:dB|%))$/i.test(volume.trim()))
+      addDiagnostic(diagnostics, source, token.start, `Unsupported <mstts:backgroundaudio volume> value "${volume}".`);
+    for (const [attribute, value] of [
+      ["fadein", attr(token, "fadein")],
+      ["fadeout", attr(token, "fadeout")]
+    ]) {
+      if (value && !isValidAzureAudioDuration(value))
+        addDiagnostic(
+          diagnostics,
+          source,
+          token.start,
+          `<mstts:backgroundaudio ${attribute}> must be a positive duration such as "500ms" or "10s".`
+        );
+    }
+    if (!token.selfClosing)
+      addDiagnostic(diagnostics, source, token.start, "<mstts:backgroundaudio> must be self-closing.");
+  }
+}
+function validateAzureSsml(ssml, options = {}) {
+  const diagnostics = [];
+  if (typeof ssml !== "string") {
+    return [
+      {
+        line: 1,
+        column: 1,
+        message: "SSML input must be a string",
+        severity: "error",
+        source: "ssml-static-validator"
+      }
+    ];
+  }
+  const maxLength = options.maxLength ?? 1e4;
+  if (ssml.length > maxLength)
+    addDiagnostic(diagnostics, ssml, maxLength, `SSML exceeds the maximum length of ${maxLength} characters.`);
+  try {
+    parseSsml(ssml);
+  } catch (error) {
+    const message = error instanceof Error ? error.message.replace(/ at position \d+$/, "") : String(error);
+    const match = / at position (\d+)$/.exec(error instanceof Error ? error.message : "");
+    addDiagnostic(diagnostics, ssml, match ? Number(match[1]) : 0, message);
+    return diagnostics;
+  }
+  const tokens = tokenizeElements(ssml);
+  const speak = tokens.find((token) => token.name.toLowerCase() === "speak");
+  const voices = tokens.filter((token) => token.name.toLowerCase() === "voice");
+  if (!speak || voices.length === 0)
+    addDiagnostic(
+      diagnostics,
+      ssml,
+      speak?.start ?? 0,
+      "Azure SSML requires at least one <voice> element under <speak>."
+    );
+  const voiceName = voices[0] ? attr(voices[0], "name") : void 0;
+  const voiceCatalog = normalizeVoiceCatalog(options);
+  const normalizeLanguage = createLanguageNormalizer(options);
+  const policySeverity = diagnosticSeverity(options.unknownVoicePolicy ?? "warn");
+  const voicesToValidate = options.validateNestedVoices === false ? voices.slice(0, 1) : voices;
+  for (const token of voicesToValidate) {
+    const name = attr(token, "name")?.trim();
+    const language = attr(token, "xml:lang")?.trim() || (speak ? attr(speak, "xml:lang")?.trim() : void 0);
+    const definition = name ? voiceCatalog.get(name.toLowerCase()) : void 0;
+    if (name && !definition && policySeverity)
+      addDiagnostic(
+        diagnostics,
+        ssml,
+        token.start,
+        `Unknown voice "${name}" is not registered in the voice catalog.`,
+        policySeverity,
+        "azure-unknown-voice"
+      );
+    if (name && language && definitionMatchesLanguage(definition, name, language, normalizeLanguage) === false)
+      addDiagnostic(
+        diagnostics,
+        ssml,
+        token.start,
+        `Voice "${name}" does not match language "${language}"; the voice name prefix indicates a different language or region.`,
+        "warning",
+        "azure-locale-mismatch"
+      );
+  }
+  for (const token of tokens) {
+    const tokenName = token.name.toLowerCase();
+    const tokenVoiceName = tokenName === "voice" ? attr(token, "name")?.trim() : tokenName === "mstts:turn" ? attr(token, "voice")?.trim() || token.parentVoiceName : options.validateNestedVoices === false ? voiceName : token.parentVoiceName;
+    validateElement(token, ssml, diagnostics, tokenVoiceName, options, voiceCatalog);
+    const definition = tokenVoiceName ? voiceCatalog.get(tokenVoiceName.toLowerCase()) : void 0;
+    validateVoiceFeatureMatrix(token, ssml, diagnostics, tokenVoiceName, definition);
+    if (tokenName === "voice" && options.model && definition?.models && !definition.models.some((model) => model.toLowerCase() === options.model?.toLowerCase())) {
+      addDiagnostic(
+        diagnostics,
+        ssml,
+        token.start,
+        `Voice "${tokenVoiceName}" does not support model "${options.model}" according to the configured feature matrix.`,
+        "error",
+        "azure-unsupported-model-for-voice"
+      );
+    }
+  }
+  return diagnostics;
+}
+var AZURE_VOICE_CATALOG_METADATA = {
+  apiVersion: "2025-10-01",
+  generatedAt: "2026-08-28T00:00:00.000Z",
+  regions: [],
+  voiceCount: AZURE_VOICE_DEFINITIONS.length
+};
 
 // packages/ssml-editor-react/src/clearSsmlDocument.ts
 function getDocumentChildren(document) {
@@ -2281,7 +2981,7 @@ function validateEntityReferences(value) {
     index = semicolon + 1;
   }
 }
-function findTagEnd2(source, start, hasInternalSubset = false) {
+function findTagEnd3(source, start, hasInternalSubset = false) {
   let quote;
   let subsetDepth = 0;
   for (let index = start + 1; index < source.length; index += 1) {
@@ -2454,7 +3154,7 @@ function parseXml(source) {
       continue;
     }
     if (source.startsWith("</", index)) {
-      const end2 = findTagEnd2(source, index);
+      const end2 = findTagEnd3(source, index);
       const name2 = parseClosingTag(source, index, end2);
       const element = currentElement(stack);
       if (element === void 0 || element.name !== name2) {
@@ -2466,7 +3166,7 @@ function parseXml(source) {
       continue;
     }
     if (source.startsWith("<!", index)) {
-      const end2 = findTagEnd2(source, index, true);
+      const end2 = findTagEnd3(source, index, true);
       parseDeclaration(source, index, end2);
       appendNode2(currentElement(stack)?.children ?? children, {
         kind: "declaration",
@@ -2475,7 +3175,7 @@ function parseXml(source) {
       index = end2 + 1;
       continue;
     }
-    const end = findTagEnd2(source, index);
+    const end = findTagEnd3(source, index);
     const { name, selfClosing } = parseStartTag(source, index, end);
     if (stack.length === 0) {
       if (root !== void 0) {
@@ -2942,6 +3642,24 @@ var SSML_COMPLETION_SNIPPETS = [
     label: "mstts:audioduration",
     insertText: '<mstts:audioduration value="10s" />'
   },
+  {
+    label: "mstts:dialog",
+    insertText: `<mstts:dialog>
+  <mstts:turn voice="\${1:en-US-JennyNeural}">\${2:text}</mstts:turn>
+</mstts:dialog>`
+  },
+  {
+    label: "mstts:turn",
+    insertText: `<mstts:turn voice="\${1:en-US-JennyNeural}">\${2:text}</mstts:turn>`
+  },
+  {
+    label: "mstts:backgroundaudio",
+    insertText: `<mstts:backgroundaudio src="\${1:https://example.com/audio.mp3}" volume="\${2:-3dB}" />`
+  },
+  {
+    label: "mstts:ttsembedding",
+    insertText: `<mstts:ttsembedding>\${1:text}</mstts:ttsembedding>`
+  },
   {
     label: "sub",
     insertText: `<sub alias="\${1:\u8AAD\u307F}">\${2:\u6F22\u5B57}</sub>`
@@ -3312,6 +4030,63 @@ var SSML_TAG_DEFINITIONS = [
         example: "10s"
       }
     ]
+  },
+  {
+    name: "mstts:dialog",
+    description: "Groups multiple Azure dialog turns that can use different voices.",
+    parameters: []
+  },
+  {
+    name: "mstts:turn",
+    description: "Adds one dialog turn using the required Azure voice name.",
+    parameters: [
+      {
+        name: "voice",
+        description: "The Azure voice used for this turn, such as `en-US-JennyNeural`.",
+        example: "en-US-JennyNeural"
+      }
+    ]
+  },
+  {
+    name: "mstts:backgroundaudio",
+    description: "Plays background audio while speech is synthesized.",
+    parameters: [
+      {
+        name: "src",
+        description: "An absolute HTTP(S) URL for the background audio file.",
+        example: "https://example.com/music.mp3"
+      },
+      {
+        name: "volume",
+        description: "The background audio volume, for example `-3dB` or `medium`.",
+        example: "-3dB"
+      },
+      {
+        name: "fadein",
+        description: "The fade-in duration, for example `1s`.",
+        example: "1s"
+      },
+      {
+        name: "fadeout",
+        description: "The fade-out duration, for example `500ms`.",
+        example: "500ms"
+      }
+    ]
+  },
+  {
+    name: "mstts:ttsembedding",
+    description: "Embeds custom voice or speaker profile metadata for Azure Speech.",
+    parameters: []
+  },
+  {
+    name: "mstts:embedding",
+    description: "Specifies embedding metadata for custom voice scenarios.",
+    parameters: []
+  },
+  {
+    name: "mstts:voiceconversion",
+    description: "Specifies voice conversion metadata for custom voice scenarios.",
+    parameters: []
   }
 ];
 var definitionsByName = /* @__PURE__ */ new Map();
@@ -3370,7 +4145,7 @@ function toRange(source, token) {
 function containsOffset(token, offset) {
   return offset >= token.start && offset < token.end;
 }
-function findTagEnd3(source, start) {
+function findTagEnd4(source, start) {
   let quote;
   for (let index = start; index < source.length; index += 1) {
     const character = source[index];
@@ -3501,7 +4276,7 @@ function findTagAtOffset(source, offset) {
       searchStart = tokenEnd2;
       continue;
     }
-    const tagEnd = findTagEnd3(source, start + 1);
+    const tagEnd = findTagEnd4(source, start + 1);
     const contentEnd = tagEnd ?? source.length;
     const tokenEnd = tagEnd === void 0 ? source.length : tagEnd + 1;
     if (offset < tokenEnd) {
@@ -3650,6 +4425,7 @@ export {
   parseSsml,
   buildPartialSsml,
   validateSsml,
+  validateAzureSsml,
   clearSsmlDocument,
   formatXmlFragment,
   getEditableRegion,
@@ -3663,4 +4439,4 @@ export {
   findSsmlHoverTarget,
   formatSsmlHover
 };
-//# sourceMappingURL=chunk-MWSDFGXW.mjs.map
+//# sourceMappingURL=chunk-JNJVTEL6.mjs.map