ssml-builder-js 2.14.0 → 2.16.0

package/dist/index.js

@@ -37,10 +37,15 @@ var __privateSet = (obj, member, value, setter) => (__accessCheck(obj, member, "
 // src/index.ts
 var src_exports = {};
 __export(src_exports, {
+  AudioFormatMismatchError: () => AudioFormatMismatchError,
   AzureTtsClient: () => AzureTtsClient,
   AzureTtsError: () => AzureTtsError,
   AzureTtsSdkError: () => AzureTtsSdkError,
   ChunkValidationError: () => ChunkValidationError,
+  DEFAULT_OUTPUT_FORMAT: () => DEFAULT_OUTPUT_FORMAT,
+  MergeError: () => MergeError,
+  SynthesisCancelledError: () => SynthesisCancelledError,
+  SynthesisTimeoutError: () => SynthesisTimeoutError,
   UnsupportedMergeFormatError: () => UnsupportedMergeFormatError,
   areAzureLanguagesEquivalent: () => areAzureLanguagesEquivalent,
   buildPartialSsml: () => buildPartialSsml,
@@ -53,6 +58,8 @@ __export(src_exports, {
   fromPlainTextToSsml: () => fromPlainTextToSsml,
   getAzureVoiceCatalogMetadata: () => getAzureVoiceCatalogMetadata,
   getBuiltInVoiceCatalogMetadata: () => getBuiltInVoiceCatalogMetadata,
+  getSsmlSourceMap: () => getSsmlSourceMap,
+  inspectAudioSpecification: () => inspectAudioSpecification,
   isValidAzureAudioDuration: () => isValidAzureAudioDuration,
   mapSsmlTextNodes: () => mapSsmlTextNodes,
   mergeAudioBuffers: () => mergeAudioBuffers,
@@ -60,6 +67,7 @@ __export(src_exports, {
   normalizeAzureLanguage: () => normalizeAzureLanguage,
   parseSsml: () => parseSsml,
   resolveMergeAudioFormat: () => resolveMergeAudioFormat,
+  resolveMimeType: () => resolveMimeType,
   splitSsmlDocument: () => splitSsmlDocument,
   synthesizeSpeech: () => synthesizeSpeech,
   synthesizeSsml: () => synthesizeSsml,
@@ -67,6 +75,7 @@ __export(src_exports, {
   synthesizeSsmlChunksSafe: () => synthesizeSsmlChunksSafe,
   synthesizeSsmlSafe: () => synthesizeSsmlSafe,
   validateAzureSsml: () => validateAzureSsml,
+  validateAzureSsmlChunks: () => validateAzureSsmlChunks,
   validateSsml: () => validateSsml,
   validateSsmlStructureIntegrity: () => validateSsmlStructureIntegrity
 });
@@ -999,6 +1008,236 @@ function buildPartialSsml(textOrOptions, context) {
   return serializePartialSsml(textOrOptions.text, textOrOptions);
 }
 
+// packages/ssml-core/src/textNodes.ts
+function decodeXmlText(value) {
+  return value.replace(/&(?:amp|apos|gt|lt|quot);|&#(?:x[\da-f]+|\d+);/gi, (entity) => {
+    if (entity === "&") return "&";
+    if (entity === "'") return "'";
+    if (entity === ">") return ">";
+    if (entity === "&lt;") return "<";
+    if (entity === "&quot;") return '"';
+    const hexadecimal = entity.toLowerCase().startsWith("&#x");
+    const digits = entity.slice(hexadecimal ? 3 : 2, -1);
+    return String.fromCodePoint(Number.parseInt(digits, hexadecimal ? 16 : 10));
+  });
+}
+function encodeXmlText(value) {
+  return value.replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;");
+}
+function decodeXmlAttribute(value) {
+  return decodeXmlText(value);
+}
+function findTagEnd(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 readTagName(tag) {
+  const match = /^<\s*([A-Za-z_][A-Za-z0-9_.:-]*)/.exec(tag);
+  return match?.[1];
+}
+function readTagAttributes(tag, name) {
+  const attributes = {};
+  const nameStart = tag.indexOf(name);
+  const attributeSource = tag.slice(nameStart + name.length, tag.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[match[1].toLowerCase()] = decodeXmlAttribute(match[3]);
+  }
+  return attributes;
+}
+function collectTextNodes(source) {
+  const nodes = [];
+  const elements = [];
+  let index = 0;
+  const addText = (start, end, rawText, sourceStart = start, sourceEnd = end) => {
+    if (!rawText) return;
+    const path = elements.map((element) => element.name);
+    const parent = elements[elements.length - 1];
+    nodes.push({
+      context: {
+        ancestorTags: path.slice(0, -1),
+        parentAttributes: { ...parent?.attributes ?? {} },
+        parentTag: parent?.name ?? "",
+        path
+      },
+      decodedText: decodeXmlText(rawText),
+      end,
+      sourceEnd,
+      sourceStart,
+      start
+    });
+  };
+  while (index < source.length) {
+    if (source[index] !== "<") {
+      const nextTag = source.indexOf("<", index);
+      const end2 = nextTag === -1 ? source.length : nextTag;
+      addText(index, end2, source.slice(index, end2));
+      index = end2;
+      continue;
+    }
+    if (source.startsWith("<!--", index)) {
+      const end2 = source.indexOf("-->", index + 4);
+      index = end2 === -1 ? source.length : end2 + 3;
+      continue;
+    }
+    if (source.startsWith("<![CDATA[", index)) {
+      const contentStart = index + 9;
+      const end2 = source.indexOf("]]>", contentStart);
+      const contentEnd = end2 === -1 ? source.length : end2;
+      addText(
+        contentStart,
+        contentEnd,
+        source.slice(contentStart, contentEnd),
+        index,
+        end2 === -1 ? source.length : end2 + 3
+      );
+      index = end2 === -1 ? source.length : end2 + 3;
+      continue;
+    }
+    if (source.startsWith("<?", index)) {
+      const end2 = source.indexOf("?>", index + 2);
+      index = end2 === -1 ? source.length : end2 + 2;
+      continue;
+    }
+    if (source.startsWith("</", index)) {
+      const end2 = findTagEnd(source, index + 2);
+      elements.pop();
+      index = end2 + 1;
+      continue;
+    }
+    const end = findTagEnd(source, index + 1);
+    const tag = source.slice(index, end + 1);
+    const name = readTagName(tag);
+    if (name && !/\/\s*>$/.test(tag)) elements.push({ attributes: readTagAttributes(tag, name), name });
+    index = end + 1;
+  }
+  return nodes;
+}
+function collectSourceMap(source) {
+  const segments = [];
+  const markers = [];
+  const elements = [];
+  let textOffset = 0;
+  let index = 0;
+  const textParts = [];
+  const addText = (value) => {
+    if (!value) return;
+    const parent = elements[elements.length - 1];
+    if (parent) parent.nextChildIndex += 1;
+    const sourceNodePath = parent?.path ?? ["speak"];
+    const start = textOffset;
+    textOffset += value.length;
+    textParts.push(value);
+    segments.push({ text: value, range: { start, end: textOffset }, sourceNodePath: [...sourceNodePath] });
+  };
+  while (index < source.length) {
+    if (source[index] !== "<") {
+      const end2 = source.indexOf("<", index);
+      const textEnd = end2 === -1 ? source.length : end2;
+      addText(decodeXmlText(source.slice(index, textEnd)));
+      index = textEnd;
+      continue;
+    }
+    if (source.startsWith("<!--", index)) {
+      const end2 = source.indexOf("-->", index + 4);
+      index = end2 === -1 ? source.length : end2 + 3;
+      continue;
+    }
+    if (source.startsWith("<![CDATA[", index)) {
+      const contentStart = index + 9;
+      const end2 = source.indexOf("]]>", contentStart);
+      const contentEnd = end2 === -1 ? source.length : end2;
+      addText(source.slice(contentStart, contentEnd));
+      index = end2 === -1 ? source.length : end2 + 3;
+      continue;
+    }
+    if (source.startsWith("<?", index)) {
+      const end2 = source.indexOf("?>", index + 2);
+      index = end2 === -1 ? source.length : end2 + 2;
+      continue;
+    }
+    const end = findTagEnd(source, index + 1);
+    const rawTag = source.slice(index, end + 1);
+    if (rawTag.startsWith("</")) {
+      elements.pop();
+      index = end + 1;
+      continue;
+    }
+    const name = readTagName(rawTag);
+    if (!name) {
+      index = end + 1;
+      continue;
+    }
+    const parent = elements[elements.length - 1];
+    const childIndex = parent?.nextChildIndex ?? 0;
+    if (parent) parent.nextChildIndex += 1;
+    const path = parent ? [...parent.path, `${name}[${childIndex}]`] : [name];
+    const attributes = readTagAttributes(rawTag, name);
+    const normalizedName = name.toLowerCase();
+    if (normalizedName === "mark" || normalizedName === "bookmark") {
+      const markerName = attributes[normalizedName === "mark" ? "name" : "mark"];
+      if (markerName) {
+        markers.push({
+          kind: normalizedName,
+          name: markerName,
+          originalTextRange: { start: textOffset, end: textOffset },
+          sourceNodePath: [...path]
+        });
+      }
+    }
+    if (!/\/\s*>$/.test(rawTag)) elements.push({ name, path, nextChildIndex: 0 });
+    index = end + 1;
+  }
+  return { text: textParts.join(""), segments, markers };
+}
+function getSsmlSourceMap(ssml) {
+  parseSsml(ssml);
+  return collectSourceMap(ssml);
+}
+function extractSsmlText(ssml) {
+  parseSsml(ssml);
+  return collectTextNodes(ssml).map((node) => node.decodedText);
+}
+async function mapSsmlTextNodes(ssml, transform, options = {}) {
+  parseSsml(ssml);
+  const nodes = collectTextNodes(ssml);
+  const skipTags = new Set((options.skipTags ?? ["phoneme", "say-as", "sub"]).map((tag) => tag.toLowerCase()));
+  const replacements = await Promise.all(
+    nodes.map(async (node) => {
+      const context = {
+        ancestorTags: [...node.context.ancestorTags],
+        parentAttributes: { ...node.context.parentAttributes },
+        parentTag: node.context.parentTag,
+        path: [...node.context.path]
+      };
+      const shouldTransform = !skipTags.has(context.parentTag.toLowerCase()) && (options.filter?.(context) ?? true);
+      if (!shouldTransform) return ssml.slice(node.sourceStart, node.sourceEnd);
+      const transformed = await transform(node.decodedText, context);
+      if (typeof transformed !== "string") {
+        throw new TypeError("SSML text node transform must return a string");
+      }
+      return transformed === node.decodedText ? ssml.slice(node.sourceStart, node.sourceEnd) : encodeXmlText(transformed);
+    })
+  );
+  let result = "";
+  let cursor = 0;
+  nodes.forEach((node, nodeIndex) => {
+    result += ssml.slice(cursor, node.sourceStart) + replacements[nodeIndex];
+    cursor = node.sourceEnd;
+  });
+  return result + ssml.slice(cursor);
+}
+
 // packages/ssml-core/src/split.ts
 var DEFAULT_MAX_LENGTH = 1e4;
 function cloneElement(element, children) {
@@ -1130,7 +1369,7 @@ function findSourceNodePath(nodes, targetOffset) {
   });
   return foundPath ?? firstPath;
 }
-function createChunk(document, nodes, chunkIndex, textStart, backgroundAudio, replicateBackgroundAudio) {
+function createChunk(document, nodes, chunkIndex, textStart, backgroundAudio, replicateBackgroundAudio, sourceMap, includeEndMarkers) {
   const chunkNodes = backgroundAudio && (replicateBackgroundAudio || chunkIndex === 0) ? [backgroundAudio, ...nodes] : nodes;
   const text = nodes.map(textFromNode).join("");
   const marks = [];
@@ -1146,7 +1385,23 @@ function createChunk(document, nodes, chunkIndex, textStart, backgroundAudio, re
     hasBackgroundAudio: chunkNodes.some(
       (node) => typeof node !== "string" && node.type !== "text" && node.type === "mstts:backgroundaudio"
     ),
-    sourceNodePath: findSourceNodePath(document.children ?? [], textStart)
+    sourceNodePath: findSourceNodePath(document.children ?? [], textStart),
+    sourceTextSegments: sourceMap.segments.filter(({ range }) => range.end > textStart && range.start < textStart + text.length).map((segment) => {
+      const start = Math.max(segment.range.start, textStart);
+      const end = Math.min(segment.range.end, textStart + text.length);
+      return {
+        text: segment.text.slice(start - segment.range.start, end - segment.range.start),
+        range: { start, end },
+        sourceNodePath: [...segment.sourceNodePath]
+      };
+    }),
+    sourceMarkers: sourceMap.markers.filter(
+      ({ originalTextRange }) => originalTextRange.start >= textStart && (originalTextRange.start < textStart + text.length || includeEndMarkers && originalTextRange.start === textStart + text.length)
+    ).map((marker) => ({
+      ...marker,
+      originalTextRange: { ...marker.originalTextRange },
+      sourceNodePath: [...marker.sourceNodePath]
+    }))
   };
 }
 function splitSsmlDocument(ssml, maxLength = DEFAULT_MAX_LENGTH, options = {}) {
@@ -1156,11 +1411,12 @@ function splitSsmlDocument(ssml, maxLength = DEFAULT_MAX_LENGTH, options = {}) {
     throw new RangeError("maxLength must be a positive integer");
   }
   const document = parseSsml(ssml);
+  const sourceMap = getSsmlSourceMap(ssml);
   const backgroundAudio = (document.children ?? []).find(
     (node) => typeof node !== "string" && node.type !== "text" && node.type === "mstts:backgroundaudio"
   );
   if (ssml.length <= resolvedMaxLength) {
-    return [createChunk(document, document.children ?? [], 0, 0, backgroundAudio, true)];
+    return [createChunk(document, document.children ?? [], 0, 0, backgroundAudio, true, sourceMap, true)];
   }
   const contentChildren = (document.children ?? []).filter((node) => node !== backgroundAudio);
   const plainDocumentLength = documentWithChildren(document, []).length;
@@ -1184,7 +1440,16 @@ function splitSsmlDocument(ssml, maxLength = DEFAULT_MAX_LENGTH, options = {}) {
   }
   if (group.length > 0) chunks.push(group);
   if (chunks.length === 0) {
-    const result = createChunk(document, [], 0, 0, backgroundAudio, resolvedOptions.replicateBackgroundAudio ?? false);
+    const result = createChunk(
+      document,
+      [],
+      0,
+      0,
+      backgroundAudio,
+      resolvedOptions.replicateBackgroundAudio ?? false,
+      sourceMap,
+      true
+    );
     if (result.ssml.length > resolvedMaxLength) {
       throw new RangeError("maxLength is too small to contain the SSML document wrapper");
     }
@@ -1198,7 +1463,9 @@ function splitSsmlDocument(ssml, maxLength = DEFAULT_MAX_LENGTH, options = {}) {
       chunkIndex,
       textStart,
       backgroundAudio,
-      resolvedOptions.replicateBackgroundAudio ?? false
+      resolvedOptions.replicateBackgroundAudio ?? false,
+      sourceMap,
+      chunkIndex === chunks.length - 1
     );
     textStart = result.originalTextRange.end;
     return result;
@@ -1221,176 +1488,27 @@ function validateSsml(xmlString) {
   }
 }
 
-// packages/ssml-core/src/textNodes.ts
-function decodeXmlText(value) {
-  return value.replace(/&(?:amp|apos|gt|lt|quot);|&#(?:x[\da-f]+|\d+);/gi, (entity) => {
-    if (entity === "&amp;") return "&";
-    if (entity === "&apos;") return "'";
-    if (entity === "&gt;") return ">";
-    if (entity === "&lt;") return "<";
-    if (entity === "&quot;") return '"';
-    const hexadecimal = entity.toLowerCase().startsWith("&#x");
-    const digits = entity.slice(hexadecimal ? 3 : 2, -1);
-    return String.fromCodePoint(Number.parseInt(digits, hexadecimal ? 16 : 10));
-  });
-}
-function encodeXmlText(value) {
-  return value.replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;");
-}
-function decodeXmlAttribute(value) {
-  return decodeXmlText(value);
-}
-function findTagEnd(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;
-    }
+// packages/ssml-core/src/migration.ts
+var DEFAULT_TRANSLATION_SKIP_TAGS = ["phoneme", "say-as", "sayAs", "sub"];
+function elementName2(element) {
+  switch (element.type) {
+    case "custom":
+    case "element":
+      return element.name;
+    case "expressAs":
+      return "mstts:express-as";
+    case "sayAs":
+      return "say-as";
+    case "silence":
+      return "mstts:silence";
+    case "viseme":
+      return "mstts:viseme";
+    default:
+      return element.type;
   }
-  return source.length - 1;
 }
-function readTagName(tag) {
-  const match = /^<\s*([A-Za-z_][A-Za-z0-9_.:-]*)/.exec(tag);
-  return match?.[1];
-}
-function readTagAttributes(tag, name) {
-  const attributes = {};
-  const nameStart = tag.indexOf(name);
-  const attributeSource = tag.slice(nameStart + name.length, tag.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[match[1].toLowerCase()] = decodeXmlAttribute(match[3]);
-  }
-  return attributes;
-}
-function collectTextNodes(source) {
-  const nodes = [];
-  const elements = [];
-  let index = 0;
-  const addText = (start, end, rawText, sourceStart = start, sourceEnd = end) => {
-    if (!rawText) return;
-    const path = elements.map((element) => element.name);
-    const parent = elements[elements.length - 1];
-    nodes.push({
-      context: {
-        ancestorTags: path.slice(0, -1),
-        parentAttributes: { ...parent?.attributes ?? {} },
-        parentTag: parent?.name ?? "",
-        path
-      },
-      decodedText: decodeXmlText(rawText),
-      end,
-      sourceEnd,
-      sourceStart,
-      start
-    });
-  };
-  while (index < source.length) {
-    if (source[index] !== "<") {
-      const nextTag = source.indexOf("<", index);
-      const end2 = nextTag === -1 ? source.length : nextTag;
-      addText(index, end2, source.slice(index, end2));
-      index = end2;
-      continue;
-    }
-    if (source.startsWith("<!--", index)) {
-      const end2 = source.indexOf("-->", index + 4);
-      index = end2 === -1 ? source.length : end2 + 3;
-      continue;
-    }
-    if (source.startsWith("<![CDATA[", index)) {
-      const contentStart = index + 9;
-      const end2 = source.indexOf("]]>", contentStart);
-      const contentEnd = end2 === -1 ? source.length : end2;
-      addText(
-        contentStart,
-        contentEnd,
-        source.slice(contentStart, contentEnd),
-        index,
-        end2 === -1 ? source.length : end2 + 3
-      );
-      index = end2 === -1 ? source.length : end2 + 3;
-      continue;
-    }
-    if (source.startsWith("<?", index)) {
-      const end2 = source.indexOf("?>", index + 2);
-      index = end2 === -1 ? source.length : end2 + 2;
-      continue;
-    }
-    if (source.startsWith("</", index)) {
-      const end2 = findTagEnd(source, index + 2);
-      elements.pop();
-      index = end2 + 1;
-      continue;
-    }
-    const end = findTagEnd(source, index + 1);
-    const tag = source.slice(index, end + 1);
-    const name = readTagName(tag);
-    if (name && !/\/\s*>$/.test(tag)) elements.push({ attributes: readTagAttributes(tag, name), name });
-    index = end + 1;
-  }
-  return nodes;
-}
-function extractSsmlText(ssml) {
-  parseSsml(ssml);
-  return collectTextNodes(ssml).map((node) => node.decodedText);
-}
-async function mapSsmlTextNodes(ssml, transform, options = {}) {
-  parseSsml(ssml);
-  const nodes = collectTextNodes(ssml);
-  const skipTags = new Set((options.skipTags ?? ["phoneme", "say-as", "sub"]).map((tag) => tag.toLowerCase()));
-  const replacements = await Promise.all(
-    nodes.map(async (node) => {
-      const context = {
-        ancestorTags: [...node.context.ancestorTags],
-        parentAttributes: { ...node.context.parentAttributes },
-        parentTag: node.context.parentTag,
-        path: [...node.context.path]
-      };
-      const shouldTransform = !skipTags.has(context.parentTag.toLowerCase()) && (options.filter?.(context) ?? true);
-      if (!shouldTransform) return ssml.slice(node.sourceStart, node.sourceEnd);
-      const transformed = await transform(node.decodedText, context);
-      if (typeof transformed !== "string") {
-        throw new TypeError("SSML text node transform must return a string");
-      }
-      return transformed === node.decodedText ? ssml.slice(node.sourceStart, node.sourceEnd) : encodeXmlText(transformed);
-    })
-  );
-  let result = "";
-  let cursor = 0;
-  nodes.forEach((node, nodeIndex) => {
-    result += ssml.slice(cursor, node.sourceStart) + replacements[nodeIndex];
-    cursor = node.sourceEnd;
-  });
-  return result + ssml.slice(cursor);
-}
-
-// packages/ssml-core/src/migration.ts
-var DEFAULT_TRANSLATION_SKIP_TAGS = ["phoneme", "say-as", "sayAs", "sub"];
-function elementName2(element) {
-  switch (element.type) {
-    case "custom":
-    case "element":
-      return element.name;
-    case "expressAs":
-      return "mstts:express-as";
-    case "sayAs":
-      return "say-as";
-    case "silence":
-      return "mstts:silence";
-    case "viseme":
-      return "mstts:viseme";
-    default:
-      return element.type;
-  }
-}
-function addAttribute2(attributes, name, value) {
-  if (value !== void 0) attributes[name] = value;
+function addAttribute2(attributes, name, value) {
+  if (value !== void 0) attributes[name] = value;
 }
 function elementAttributes(element) {
   const attributes = { ...element.attributes ?? {} };
@@ -1790,34 +1908,51 @@ function createAzureUrlValidatorRunner(validator, options = {}) {
   const inFlight = /* @__PURE__ */ new Map();
   const waiters = [];
   let active = 0;
-  const acquire = async () => {
+  const configuredSignal = options.signal ?? new AbortController().signal;
+  const acquire = async (signal) => {
+    if (signal.aborted) throw new Error("URL validation was aborted.");
     if (active < concurrency) {
       active += 1;
       return;
     }
-    await new Promise((resolve) => waiters.push(resolve));
+    await new Promise((resolve, reject) => {
+      let waiter;
+      const abortHandler = () => {
+        const index = waiters.indexOf(waiter);
+        if (index >= 0) waiters.splice(index, 1);
+        signal.removeEventListener("abort", abortHandler);
+        reject(new Error("URL validation was aborted."));
+      };
+      signal.addEventListener("abort", abortHandler, { once: true });
+      waiter = () => {
+        signal.removeEventListener("abort", abortHandler);
+        resolve();
+      };
+      waiters.push(waiter);
+    });
     active += 1;
   };
   const release = () => {
     active -= 1;
     waiters.shift()?.();
   };
-  const check = async (url, context) => {
-    if (options.signal?.aborted) throw new Error("URL validation was aborted.");
-    const cached = cache.get(url);
+  const check = async (url, context, signal = configuredSignal) => {
+    if (signal.aborted) throw new Error("URL validation was aborted.");
+    const key = `${context.tag}:${context.attribute}:${url}`;
+    const cached = cache.get(key);
     if (cached !== void 0) return cached;
-    const existing = inFlight.get(url);
+    const existing = inFlight.get(key);
     if (existing) return existing;
     const promise = (async () => {
-      await acquire();
+      await acquire(signal);
       try {
-        if (options.signal?.aborted) throw new Error("URL validation was aborted.");
-        const validation = Promise.resolve(validator(url, context));
+        if (signal.aborted) throw new Error("URL validation was aborted.");
+        const validation = Promise.resolve(validator(url, context, signal));
         let timer;
         let abortHandler;
         const cancellation = new Promise((_resolve, reject) => {
           abortHandler = () => reject(new Error("URL validation was aborted."));
-          options.signal?.addEventListener("abort", abortHandler, { once: true });
+          signal.addEventListener("abort", abortHandler, { once: true });
           if (options.timeoutMs !== void 0 && options.timeoutMs > 0) {
             timer = setTimeout(
               () => reject(new Error(`URL validation timed out after ${options.timeoutMs} ms.`)),
@@ -1827,24 +1962,24 @@ function createAzureUrlValidatorRunner(validator, options = {}) {
         });
         try {
           const result = await (timer || abortHandler ? Promise.race([validation, cancellation]) : validation);
-          cache.set(url, result);
+          cache.set(key, result);
           return result;
         } finally {
           if (timer) clearTimeout(timer);
-          if (abortHandler) options.signal?.removeEventListener("abort", abortHandler);
+          if (abortHandler) signal.removeEventListener("abort", abortHandler);
         }
       } finally {
         release();
       }
     })();
-    inFlight.set(url, promise);
+    inFlight.set(key, promise);
     try {
       return await promise;
     } finally {
-      inFlight.delete(url);
+      inFlight.delete(key);
     }
   };
-  return (url, context) => check(url, context);
+  return (url, context, signal) => check(url, context, signal ?? configuredSignal);
 }
 var ALLOWED_BREAK_STRENGTHS = /* @__PURE__ */ new Set(["none", "x-weak", "weak", "medium", "strong", "x-strong"]);
 var ALLOWED_SAY_AS = /* @__PURE__ */ new Set([
@@ -1957,6 +2092,7 @@ function tokenizeElements(source) {
     if (parent) parent.childElementCount += 1;
     const parentVoiceName = [...openElements].reverse().find((element) => element.voiceName)?.voiceName;
     const tokenName = nameMatch[1];
+    const path = parent ? [...parent.path, `${tokenName}[${childElementIndex ?? 0}]`] : [tokenName];
     const tokenVoiceName = tokenName.toLowerCase() === "voice" ? attributes.get("name") : tokenName.toLowerCase() === "mstts:turn" ? attributes.get("voice") ?? parentVoiceName : parentVoiceName;
     tokens.push({
       attributes,
@@ -1967,12 +2103,14 @@ function tokenizeElements(source) {
       parentName: parent?.name,
       parentVoiceName,
       selfClosing,
-      start
+      start,
+      path
     });
     if (!selfClosing) {
       openElements.push({
         childElementCount: 0,
         name: tokenName,
+        path,
         voiceName: tokenVoiceName
       });
     }
@@ -1985,13 +2123,14 @@ function location(source, 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", code) {
+function addDiagnostic(diagnostics, source, offset, message, severity = "error", code, metadata = {}) {
   diagnostics.push({
     ...location(source, offset),
     message,
     severity,
     source: "ssml-static-validator",
-    ...code ? { code } : {}
+    ...code ? { code } : {},
+    ...metadata
   });
 }
 function isSupportedProsodyRate(value) {
@@ -2464,9 +2603,11 @@ function validateAzureSsmlStatic(ssml, options = {}) {
   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;
+    const tokenDiagnosticStart = diagnostics.length;
     validateElement(token, ssml, diagnostics, tokenVoiceName, options, voiceCatalog);
     const definition = tokenVoiceName ? voiceCatalog.get(tokenVoiceName.toLowerCase()) : void 0;
     validateVoiceFeatureMatrix(token, ssml, diagnostics, tokenVoiceName, definition);
+    annotateTokenDiagnostics(diagnostics, tokenDiagnosticStart, token, options, tokenVoiceName);
     if (tokenName === "voice" && options.model && definition?.models && !definition.models.some((model) => model.toLowerCase() === options.model?.toLowerCase())) {
       addDiagnostic(
         diagnostics,
@@ -2488,18 +2629,37 @@ function urlAttributes(token) {
     return value === void 0 ? [] : [{ attribute, value }];
   });
 }
+function annotateTokenDiagnostics(diagnostics, startIndex, token, options, voiceName) {
+  const attributes = [...token.attributes.keys()];
+  for (const diagnostic of diagnostics.slice(startIndex)) {
+    const attributeName = attributes.find(
+      (attribute) => new RegExp(`(?:<[^> ]+\\s+|")${attribute}(?:"|>|\\s)`, "i").test(diagnostic.message)
+    );
+    const nodePath = options.sourceNodePath ? [...options.sourceNodePath] : [...token.path];
+    Object.assign(diagnostic, {
+      range: { start: token.start, end: token.end + 1 },
+      tagName: token.name,
+      ...attributeName ? { attributeName } : {},
+      ...voiceName ? { voiceName } : {},
+      ...options.chunkIndex !== void 0 ? { chunkIndex: options.chunkIndex } : {},
+      nodePath,
+      targetNodePath: [...token.path]
+    });
+  }
+}
 function validateAzureSsml(ssml, options = {}) {
   const diagnostics = validateAzureSsmlStatic(ssml, options);
   const validator = options.urlValidator ?? options.customUrlValidator;
   if (!validator || typeof ssml !== "string") return diagnostics;
   const runnerOptions = options.urlValidation ?? {};
-  const boundedValidator = createAzureUrlValidatorRunner(validator, {
+  const boundedValidator = options.urlValidatorRunner ?? createAzureUrlValidatorRunner(validator, {
     ...runnerOptions,
     ...options.urlValidatorConcurrency !== void 0 ? { concurrency: options.urlValidatorConcurrency } : {},
     ...options.urlValidatorTimeoutMs !== void 0 ? { timeoutMs: options.urlValidatorTimeoutMs } : {},
     ...options.urlValidatorSignal ? { signal: options.urlValidatorSignal } : {},
     ...options.urlValidatorCache ? { cache: options.urlValidatorCache } : {}
   });
+  const validationSignal = options.urlValidatorSignal ?? options.urlValidation?.signal ?? new AbortController().signal;
   let tokens;
   try {
     tokens = tokenizeElements(ssml);
@@ -2509,30 +2669,54 @@ function validateAzureSsml(ssml, options = {}) {
   const checks = tokens.flatMap(
     (token) => urlAttributes(token).map(async ({ attribute, value }) => {
       try {
-        const result = await boundedValidator(value, { tag: token.name, attribute });
+        const result = await boundedValidator(
+          value,
+          { tag: token.name, attribute, ...options.sourceNodePath ? { sourceNodePath: options.sourceNodePath } : {} },
+          validationSignal
+        );
         const valid = typeof result === "boolean" ? result : result.valid;
         if (!valid) {
           const reason = typeof result === "boolean" ? void 0 : result.reason;
+          const diagnosticStart = diagnostics.length;
           addDiagnostic(
             diagnostics,
             ssml,
             token.start,
             `<${token.name} ${attribute}> was rejected by the custom URL validator${reason ? `: ${reason}` : "."}`
           );
+          annotateTokenDiagnostics(diagnostics, diagnosticStart, token, options, token.parentVoiceName);
         }
       } catch (error) {
         const reason = error instanceof Error ? error.message : String(error);
+        const diagnosticStart = diagnostics.length;
         addDiagnostic(
           diagnostics,
           ssml,
           token.start,
           `<${token.name} ${attribute}> could not be validated by the custom URL validator: ${reason}`
         );
+        annotateTokenDiagnostics(diagnostics, diagnosticStart, token, options, token.parentVoiceName);
       }
     })
   );
   return Promise.all(checks).then(() => diagnostics);
 }
+async function validateAzureSsmlChunks(chunks, options = {}) {
+  const validator = options.urlValidator ?? options.customUrlValidator;
+  const sharedOptions = validator ? {
+    ...options,
+    urlValidatorRunner: options.urlValidatorRunner ?? createAzureUrlValidatorRunner(validator, {
+      ...options.urlValidation ?? {},
+      ...options.urlValidatorConcurrency !== void 0 ? { concurrency: options.urlValidatorConcurrency } : {},
+      ...options.urlValidatorTimeoutMs !== void 0 ? { timeoutMs: options.urlValidatorTimeoutMs } : {},
+      ...options.urlValidatorSignal ? { signal: options.urlValidatorSignal } : {},
+      ...options.urlValidatorCache ? { cache: options.urlValidatorCache } : {}
+    })
+  } : options;
+  return Promise.all(
+    chunks.map((chunk, chunkIndex) => Promise.resolve(validateAzureSsml(chunk, { ...sharedOptions, chunkIndex })))
+  );
+}
 
 // packages/ssml-core/src/generated/azureVoiceCatalog.ts
 var AZURE_VOICE_CATALOG_METADATA = {
@@ -2555,6 +2739,7 @@ var getBuiltInVoiceCatalogMetadata = getAzureVoiceCatalogMetadata;
 var AzureTtsError = class extends Error {
   constructor(status, statusText, responseBody, requestId) {
     super(`Azure TTS request failed: ${status} ${statusText}`);
+    this.kind = "azure-api-error";
     this.name = "AzureTtsError";
     this.status = status;
     this.statusText = statusText;
@@ -2570,13 +2755,52 @@ var AzureTtsSdkError = class extends AzureTtsError {
     this.errorDetails = errorDetails;
   }
 };
+var SynthesisCancelledError = class extends Error {
+  constructor(message = "Speech synthesis was cancelled.") {
+    super(message);
+    this.kind = "cancelled";
+    this.name = "SynthesisCancelledError";
+  }
+};
+var SynthesisTimeoutError = class extends Error {
+  constructor(message) {
+    super(message);
+    this.kind = "timeout";
+    this.name = "SynthesisTimeoutError";
+  }
+};
+var MergeError = class extends Error {
+  constructor(message, cause) {
+    super(message);
+    this.kind = "merge-error";
+    this.name = "MergeError";
+    this.cause = cause;
+  }
+};
+var AudioFormatMismatchError = class extends Error {
+  constructor(message, inputSpecs = []) {
+    super(message);
+    this.kind = "audio-format-mismatch";
+    this.name = "AudioFormatMismatchError";
+    this.inputSpecs = inputSpecs;
+  }
+};
 var UnsupportedMergeFormatError = class extends Error {
   constructor(format) {
     super(`Audio format "${format}" cannot be safely concatenated; container re-multiplexing is required.`);
+    this.kind = "unsupported-format-error";
     this.name = "UnsupportedMergeFormatError";
     this.format = format;
   }
 };
+function toSynthesisError(error) {
+  if (error instanceof AzureTtsError || error instanceof MergeError || error instanceof AudioFormatMismatchError || error instanceof UnsupportedMergeFormatError || error instanceof SynthesisCancelledError || error instanceof SynthesisTimeoutError)
+    return error;
+  const message = error instanceof Error ? error.message : String(error);
+  if (/cancel|abort/i.test(message)) return new SynthesisCancelledError(message);
+  if (/tim(?:e|ed) ?out/i.test(message)) return new SynthesisTimeoutError(message);
+  return createSpeechSdkError(error);
+}
 function createSpeechSdkError(error) {
   const message = error instanceof Error ? error.message : String(error);
   return new AzureTtsSdkError(message);
@@ -2585,657 +2809,191 @@ function createSpeechSdkError(error) {
 // packages/azure-tts-client/src/synthesis.ts
 var SpeechSDK2 = __toESM(require("microsoft-cognitiveservices-speech-sdk"));
 
-// packages/azure-tts-client/src/speechConfig.ts
-var import_microsoft_cognitiveservices_speech_sdk = require("microsoft-cognitiveservices-speech-sdk");
-
-// packages/azure-tts-client/src/outputFormats.ts
-var SpeechSDK = __toESM(require("microsoft-cognitiveservices-speech-sdk"));
-var DEFAULT_OUTPUT_FORMAT = "audio-16khz-128kbitrate-mono-mp3";
-var OUTPUT_FORMATS = {
-  "raw-8khz-8bit-mono-mulaw": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz8BitMonoMULaw,
-  "riff-16khz-16kbps-mono-siren": SpeechSDK.SpeechSynthesisOutputFormat.Riff16Khz16KbpsMonoSiren,
-  "audio-16khz-16kbps-mono-siren": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz16KbpsMonoSiren,
-  "audio-16khz-32kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz32KBitRateMonoMp3,
-  "audio-16khz-128kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz128KBitRateMonoMp3,
-  "audio-16khz-64kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz64KBitRateMonoMp3,
-  "audio-24khz-48kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz48KBitRateMonoMp3,
-  "audio-24khz-96kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz96KBitRateMonoMp3,
-  "audio-24khz-160kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz160KBitRateMonoMp3,
-  "raw-16khz-16bit-mono-truesilk": SpeechSDK.SpeechSynthesisOutputFormat.Raw16Khz16BitMonoTrueSilk,
-  "riff-16khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Riff16Khz16BitMonoPcm,
-  "riff-8khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz16BitMonoPcm,
-  "riff-24khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Riff24Khz16BitMonoPcm,
-  "riff-8khz-8bit-mono-mulaw": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz8BitMonoMULaw,
-  "raw-16khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Raw16Khz16BitMonoPcm,
-  "raw-24khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Raw24Khz16BitMonoPcm,
-  "raw-8khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz16BitMonoPcm,
-  "ogg-16khz-16bit-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Ogg16Khz16BitMonoOpus,
-  "ogg-24khz-16bit-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Ogg24Khz16BitMonoOpus,
-  "raw-48khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Raw48Khz16BitMonoPcm,
-  "riff-48khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Riff48Khz16BitMonoPcm,
-  "audio-48khz-96kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio48Khz96KBitRateMonoMp3,
-  "audio-48khz-192kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio48Khz192KBitRateMonoMp3,
-  "ogg-48khz-16bit-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Ogg48Khz16BitMonoOpus,
-  "webm-16khz-16bit-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Webm16Khz16BitMonoOpus,
-  "webm-24khz-16bit-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Webm24Khz16BitMonoOpus,
-  "webm-24khz-16bit-24kbps-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Webm24Khz16Bit24KbpsMonoOpus,
-  "raw-24khz-16bit-mono-truesilk": SpeechSDK.SpeechSynthesisOutputFormat.Raw24Khz16BitMonoTrueSilk,
-  "raw-8khz-8bit-mono-alaw": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz8BitMonoALaw,
-  "riff-8khz-8bit-mono-alaw": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz8BitMonoALaw,
-  "audio-16khz-16bit-32kbps-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz16Bit32KbpsMonoOpus,
-  "audio-24khz-16bit-48kbps-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz16Bit48KbpsMonoOpus,
-  "audio-24khz-16bit-24kbps-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz16Bit24KbpsMonoOpus,
-  "raw-22050hz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Raw22050Hz16BitMonoPcm,
-  "riff-22050hz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Riff22050Hz16BitMonoPcm,
-  "raw-44100hz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Raw44100Hz16BitMonoPcm,
-  "riff-44100hz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Riff44100Hz16BitMonoPcm,
-  "amr-wb-16000hz": SpeechSDK.SpeechSynthesisOutputFormat.AmrWb16000Hz,
-  "g722-16khz-64kbps": SpeechSDK.SpeechSynthesisOutputFormat.G72216Khz64Kbps
+// packages/ssml-core/dist/index.mjs
+var __typeError2 = (msg) => {
+  throw TypeError(msg);
 };
-function resolveOutputFormat(outputFormat) {
-  const resolvedFormat = OUTPUT_FORMATS[outputFormat];
-  if (resolvedFormat === void 0) {
-    throw new Error(`Unsupported Azure Speech output format: ${outputFormat}`);
-  }
-  return resolvedFormat;
-}
-
-// packages/azure-tts-client/src/speechConfig.ts
-function resolveEndpoint(config) {
-  const endpoint = config.endpoint?.trim() || "https://{region}.tts.speech.microsoft.com/cognitiveservices/v1";
-  return endpoint.replace(/\{region\}/g, encodeURIComponent(config.region));
-}
-function createSpeechConfig(config) {
-  const { outputFormat = DEFAULT_OUTPUT_FORMAT, subscriptionKey } = config;
-  const endpoint = new URL(resolveEndpoint(config));
-  const speechConfig = import_microsoft_cognitiveservices_speech_sdk.SpeechConfig.fromEndpoint(endpoint, subscriptionKey);
-  speechConfig.speechSynthesisOutputFormat = resolveOutputFormat(outputFormat);
-  return speechConfig;
-}
-
-// packages/azure-tts-client/src/synthesis.ts
-function ascii(bytes, offset, value) {
-  return [...value].every((character, index) => bytes[offset + index] === character.charCodeAt(0));
-}
-function readUint32(bytes, offset) {
-  return new DataView(bytes.buffer, bytes.byteOffset, bytes.byteLength).getUint32(offset, true);
-}
-function parseWav(buffer) {
-  const bytes = new Uint8Array(buffer);
-  if (bytes.byteLength < 12 || !ascii(bytes, 0, "RIFF") || !ascii(bytes, 8, "WAVE")) {
-    throw new Error("Invalid WAV/RIFF audio buffer.");
-  }
-  const chunks = [];
-  const dataParts = [];
-  let format;
-  let offset = 12;
-  while (offset < bytes.byteLength) {
-    if (offset + 8 > bytes.byteLength) throw new Error("Invalid WAV chunk header.");
-    const id = String.fromCharCode(...bytes.slice(offset, offset + 4));
-    const size = readUint32(bytes, offset + 4);
-    const dataStart = offset + 8;
-    const dataEnd = dataStart + size;
-    if (dataEnd > bytes.byteLength) throw new Error(`WAV chunk "${id}" exceeds the audio buffer.`);
-    const data2 = bytes.slice(dataStart, dataEnd);
-    chunks.push({ id, data: data2 });
-    if (id === "fmt ") format ?? (format = data2);
-    if (id === "data") dataParts.push(data2);
-    offset = dataEnd + (size & 1);
-    if (offset > bytes.byteLength) throw new Error("Invalid WAV chunk padding.");
-  }
-  if (!format || dataParts.length === 0) throw new Error("WAV audio must contain fmt and data chunks.");
-  const dataLength = dataParts.reduce((total, part) => total + part.byteLength, 0);
-  const data = new Uint8Array(dataLength);
-  let dataOffset = 0;
-  for (const part of dataParts) {
-    data.set(part, dataOffset);
-    dataOffset += part.byteLength;
-  }
-  return { chunks, data, format };
+var __accessCheck2 = (obj, member, msg) => member.has(obj) || __typeError2("Cannot " + msg);
+var __privateGet2 = (obj, member, getter) => (__accessCheck2(obj, member, "read from private field"), getter ? getter.call(obj) : member.get(obj));
+var __privateAdd2 = (obj, member, value) => member.has(obj) ? __typeError2("Cannot add the same private member more than once") : member instanceof WeakSet ? member.add(obj) : member.set(obj, value);
+var __privateSet2 = (obj, member, value, setter) => (__accessCheck2(obj, member, "write to private field"), setter ? setter.call(obj, value) : member.set(obj, value), value);
+var SYNTHESIS_NAMESPACE2 = "http://www.w3.org/2001/10/synthesis";
+var MSTTS_NAMESPACE2 = "http://www.w3.org/2001/mstts";
+var MAX_NESTING_DEPTH2 = 1e3;
+var SSML_TAGS2 = {
+  SPEAK: "speak",
+  VOICE: "voice",
+  PROSODY: "prosody",
+  BREAK: "break",
+  EXPRESS_AS: "express-as",
+  EXPRESS_AS_CAMEL: "expressAs",
+  MSTTS_EXPRESS_AS: "mstts:express-as",
+  SAY_AS: "say-as",
+  SAY_AS_CAMEL: "sayAs",
+  PHONEME: "phoneme",
+  EMPHASIS: "emphasis",
+  AUDIO: "audio",
+  SUB: "sub",
+  LANG: "lang",
+  MARK: "mark",
+  BOOKMARK: "bookmark",
+  LEXICON: "lexicon",
+  PARAGRAPH: "p",
+  SENTENCE: "s",
+  WORD: "w",
+  MSTTS_SILENCE: "mstts:silence",
+  SILENCE: "silence",
+  MSTTS_VISEME: "mstts:viseme",
+  VISEME: "viseme",
+  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_ATTRS2 = {
+  VERSION: "version",
+  XMLNS: "xmlns",
+  XML_LANG: "xml:lang",
+  LANG: "lang",
+  MSTTS_XMLNS: "xmlns:mstts",
+  NAME: "name",
+  VOICE: "voice",
+  SPEAKER: "speaker",
+  EFFECT: "effect",
+  RATE: "rate",
+  PITCH: "pitch",
+  VOLUME: "volume",
+  CONTOUR: "contour",
+  RANGE: "range",
+  TIME: "time",
+  STRENGTH: "strength",
+  STYLE: "style",
+  STYLE_DEGREE: "styledegree",
+  STYLE_DEGREE_CAMEL: "styleDegree",
+  STYLE_DEGREE_HYPHEN: "style-degree",
+  ROLE: "role",
+  INTERPRET_AS: "interpret-as",
+  FORMAT: "format",
+  DETAIL: "detail",
+  ALPHABET: "alphabet",
+  PH: "ph",
+  LEVEL: "level",
+  SRC: "src",
+  DESC: "desc",
+  CLIP_BEGIN: "clipBegin",
+  CLIP_END: "clipEnd",
+  SPEED: "speed",
+  REPEAT_COUNT: "repeatCount",
+  REPEAT_DURATION: "repeatDuration",
+  SOUND_LEVEL: "soundLevel",
+  ALIAS: "alias",
+  MARK: "mark",
+  URI: "uri",
+  ID: "id",
+  MODEL: "model",
+  PROFILE: "profile",
+  URL: "url",
+  SPEAKER_PROFILE_ID: "speakerProfileId",
+  TYPE: "type",
+  VALUE: "value",
+  FADE_IN: "fadein",
+  FADE_OUT: "fadeout"
+};
+var XML_ENTITIES2 = {
+  amp: "&",
+  apos: "'",
+  gt: ">",
+  lt: "<",
+  quot: '"'
+};
+function hasOwn2(object, property) {
+  return Object.getOwnPropertyDescriptor(object, property) !== void 0;
 }
-function writeUint32(target, offset, value) {
-  new DataView(target.buffer).setUint32(offset, value, true);
+function setAttribute2(attributes, name, value) {
+  Object.defineProperty(attributes, name, {
+    configurable: true,
+    enumerable: true,
+    value,
+    writable: true
+  });
 }
-function writeChunk(target, offset, id, data) {
-  for (let index = 0; index < 4; index += 1) target[offset + index] = id.charCodeAt(index) ?? 0;
-  writeUint32(target, offset + 4, data.byteLength);
-  target.set(data, offset + 8);
-  const end = offset + 8 + data.byteLength;
-  if (data.byteLength & 1) target[end] = 0;
-  return end + (data.byteLength & 1);
+function decodeEntity2(entity) {
+  const namedValue = hasOwn2(XML_ENTITIES2, entity) ? XML_ENTITIES2[entity] : void 0;
+  if (namedValue !== void 0) {
+    return namedValue;
+  }
+  const isHexadecimal = entity.startsWith("#x") || entity.startsWith("#X");
+  const isDecimal = entity.startsWith("#");
+  if (!isHexadecimal && !isDecimal) {
+    throw new Error(`Unknown XML entity: &${entity};`);
+  }
+  const digits = entity.slice(isHexadecimal ? 2 : 1);
+  const codePoint = Number.parseInt(digits, isHexadecimal ? 16 : 10);
+  if (!digits || !Number.isInteger(codePoint) || codePoint < 0 || codePoint > 1114111 || codePoint >= 55296 && codePoint <= 57343 || codePoint < 32 && ![9, 10, 13].includes(codePoint)) {
+    throw new Error(`Invalid XML character reference: &${entity};`);
+  }
+  return String.fromCodePoint(codePoint);
 }
-function mergeWavBuffers(buffers) {
-  if (buffers.length === 0) return new ArrayBuffer(0);
-  const parsed = buffers.map(parseWav);
-  const first = parsed[0];
-  if (!first) throw new Error("At least one WAV buffer is required.");
-  if (parsed.some(
-    (item) => item.format.length !== first.format.length || item.format.some((value, i) => value !== first.format[i])
-  ))
-    throw new Error("WAV buffers have incompatible fmt chunks.");
-  const dataLength = parsed.reduce((total, item) => total + item.data.byteLength, 0);
-  const nonDataLength = first.chunks.reduce(
-    (total, chunk) => chunk.id === "data" ? total : total + 8 + chunk.data.byteLength + (chunk.data.byteLength & 1),
-    0
-  );
-  const outputLength = 12 + nonDataLength + 8 + dataLength + (dataLength & 1);
-  if (outputLength - 8 > 4294967295) throw new RangeError("Merged WAV exceeds the RIFF format size limit.");
-  const output = new Uint8Array(outputLength);
-  output.set(Uint8Array.from([82, 73, 70, 70]), 0);
-  writeUint32(output, 4, outputLength - 8);
-  output.set(Uint8Array.from([87, 65, 86, 69]), 8);
-  let outputOffset = 12;
-  let dataWritten = false;
-  for (const chunk of first.chunks) {
-    if (chunk.id === "data") {
-      if (dataWritten) continue;
-      const data = new Uint8Array(dataLength);
-      let dataOffset = 0;
-      for (const item of parsed) {
-        data.set(item.data, dataOffset);
-        dataOffset += item.data.byteLength;
-      }
-      outputOffset = writeChunk(output, outputOffset, "data", data);
-      dataWritten = true;
-    } else {
-      outputOffset = writeChunk(output, outputOffset, chunk.id, chunk.data);
+function decodeXmlEntities2(value) {
+  let result = "";
+  let start = 0;
+  while (true) {
+    const ampersand = value.indexOf("&", start);
+    if (ampersand === -1) {
+      return result + value.slice(start);
     }
+    result += value.slice(start, ampersand);
+    const semicolon = value.indexOf(";", ampersand + 1);
+    if (semicolon === -1) {
+      throw new Error("Unterminated XML entity reference");
+    }
+    result += decodeEntity2(value.slice(ampersand + 1, semicolon));
+    start = semicolon + 1;
   }
-  if (!dataWritten) throw new Error("WAV audio must contain a data chunk.");
-  return output.buffer;
-}
-function skipId3v2(bytes) {
-  if (!ascii(bytes, 0, "ID3") || bytes.byteLength < 10) return 0;
-  const size = [bytes[6], bytes[7], bytes[8], bytes[9]].reduce((total, value) => total << 7 | value & 127, 0);
-  const hasFooter = (bytes[5] & 16) !== 0;
-  return Math.min(bytes.byteLength, 10 + size + (hasFooter ? 10 : 0));
-}
-function stripMp3Tags(buffer) {
-  const bytes = new Uint8Array(buffer);
-  const start = skipId3v2(bytes);
-  const end = bytes.byteLength >= 128 && ascii(bytes, bytes.byteLength - 128, "TAG") ? bytes.byteLength - 128 : bytes.byteLength;
-  return bytes.slice(Math.min(start, end), end);
-}
-function isMp3Format(format) {
-  return /(?:mp3|mpeg)/i.test(format);
 }
-function isWavFormat(format) {
-  return /(?:wav|wave|riff)/i.test(format);
-}
-function isRawFormat(format) {
-  return /^raw(?:-|$)/i.test(format);
+function isXmlNameStart2(value) {
+  return value !== void 0 && /[A-Za-z_]/.test(value);
 }
-function resolveMergeAudioFormat(format) {
-  if (isWavFormat(format)) return "wav";
-  if (isMp3Format(format)) return "mp3";
-  if (isRawFormat(format)) return "raw";
-  return void 0;
+function isXmlNameCharacter2(value) {
+  return value !== void 0 && /[A-Za-z0-9_.:-]/.test(value);
 }
-function canMergeAudioFormat(format) {
-  return resolveMergeAudioFormat(format) !== void 0;
+function isXmlWhitespace2(value) {
+  return value === " " || value === "	" || value === "\r" || value === "\n";
 }
-function mergeAudioBuffers(buffers, format) {
-  if (isWavFormat(format)) return mergeWavBuffers(buffers);
-  if (isMp3Format(format)) {
-    const parts = buffers.map(stripMp3Tags);
-    const output = new Uint8Array(parts.reduce((total, part) => total + part.byteLength, 0));
-    let offset = 0;
-    for (const part of parts) {
-      output.set(part, offset);
-      offset += part.byteLength;
-    }
-    return output.buffer;
+function removeStandardNamespaceAttributes2(attributes) {
+  if (attributes[SSML_ATTRS2.XMLNS] === SYNTHESIS_NAMESPACE2) {
+    delete attributes[SSML_ATTRS2.XMLNS];
   }
-  if (isRawFormat(format)) {
-    const output = new Uint8Array(buffers.reduce((total, buffer) => total + buffer.byteLength, 0));
-    let offset = 0;
-    for (const buffer of buffers) {
-      output.set(new Uint8Array(buffer), offset);
-      offset += buffer.byteLength;
-    }
-    return output.buffer;
+  if (attributes[SSML_ATTRS2.MSTTS_XMLNS] === MSTTS_NAMESPACE2) {
+    delete attributes[SSML_ATTRS2.MSTTS_XMLNS];
   }
-  throw new UnsupportedMergeFormatError(format);
 }
-function closeSpeechResources(speechConfig, synthesizer) {
-  try {
-    synthesizer.close();
-  } catch {
+var _index2;
+var XmlParser2 = class {
+  constructor(source) {
+    __privateAdd2(this, _index2, 0);
+    this.source = source;
   }
-  try {
-    speechConfig.close();
-  } catch {
-  }
-}
-var ticksToMilliseconds = (ticks) => Math.max(0, ticks) / 1e4;
-async function synthesizeSsml(ssml, config) {
-  if (config.signal?.aborted) {
-    throw createSpeechSdkError("Speech synthesis was cancelled.");
-  }
-  const speechConfig = createSpeechConfig(config);
-  const synthesizer = new SpeechSDK2.SpeechSynthesizer(speechConfig, null);
-  return await new Promise((resolve, reject) => {
-    let resourcesClosed = false;
-    let settled = false;
-    let timeout;
-    let abortHandler;
-    const cleanup = () => {
-      if (timeout) clearTimeout(timeout);
-      if (abortHandler) config.signal?.removeEventListener("abort", abortHandler);
-    };
-    const closeResources = () => {
-      if (resourcesClosed) return;
-      resourcesClosed = true;
-      closeSpeechResources(speechConfig, synthesizer);
-    };
-    const rejectWithError = (error) => {
-      if (settled) return;
-      settled = true;
-      cleanup();
-      closeResources();
-      reject(createSpeechSdkError(error));
-    };
-    const boundaries = [];
-    const visemes = [];
-    const bookmarks = [];
-    synthesizer.wordBoundary = (_sender, event) => {
-      boundaries.push({
-        text: event.text,
-        audioOffsetMs: ticksToMilliseconds(event.audioOffset),
-        durationMs: ticksToMilliseconds(event.duration)
-      });
-    };
-    synthesizer.visemeReceived = (_sender, event) => {
-      visemes.push({ visemeId: event.visemeId, audioOffsetMs: ticksToMilliseconds(event.audioOffset) });
-    };
-    synthesizer.bookmarkReached = (_sender, event) => {
-      bookmarks.push({ name: event.text, audioOffsetMs: ticksToMilliseconds(event.audioOffset) });
-    };
-    const cb = (result) => {
-      if (settled) return;
-      const { reason, errorDetails } = result;
-      if (reason !== SpeechSDK2.ResultReason.SynthesizingAudioCompleted) {
-        const err = errorDetails || `Speech synthesis failed with reason ${reason}.`;
-        rejectWithError(err);
-        return;
-      }
-      settled = true;
-      cleanup();
-      closeResources();
-      const eventDurationMs = Math.max(
-        0,
-        ...(boundaries ?? []).map((boundary) => boundary.audioOffsetMs + boundary.durationMs),
-        ...(visemes ?? []).map((viseme) => viseme.audioOffsetMs),
-        ...(bookmarks ?? []).map((bookmark) => bookmark.audioOffsetMs)
-      );
-      const durationMs = result.audioDuration ? ticksToMilliseconds(result.audioDuration) : eventDurationMs;
-      const requestId = result.resultId;
-      const addSourceMetadata = (event) => ({
-        ...event,
-        ...config.sourceTextRange ? { textRange: { ...config.sourceTextRange } } : {},
-        ...config.sourceTextRange ? { originalTextRange: { ...config.sourceTextRange } } : {},
-        ...config.chunkIndex !== void 0 ? { chunkIndex: config.chunkIndex } : {},
-        ...config.sourceNodePath ? { sourceNodePath: [...config.sourceNodePath] } : {},
-        ...requestId ? { requestId } : {}
-      });
-      const sourceBoundaries = boundaries.map((boundary) => addSourceMetadata(boundary));
-      const sourceVisemes = visemes.map((viseme) => addSourceMetadata(viseme));
-      const sourceBookmarks = bookmarks.map((bookmark) => addSourceMetadata(bookmark));
-      resolve({
-        audioData: result.audioData,
-        durationMs,
-        ...config.sourceTextRange ? { textRange: { ...config.sourceTextRange } } : {},
-        ...requestId ? { requestId } : {},
-        ...sourceBoundaries.length > 0 ? { boundaries: sourceBoundaries, wordBoundary: sourceBoundaries, wordBoundaries: sourceBoundaries } : {},
-        ...sourceVisemes.length > 0 ? { visemes: sourceVisemes } : {},
-        ...sourceBookmarks.length > 0 ? { bookmarks: sourceBookmarks } : {}
-      });
-    };
-    try {
-      if (config.signal) {
-        abortHandler = () => rejectWithError("Speech synthesis was cancelled.");
-        config.signal.addEventListener("abort", abortHandler, { once: true });
-      }
-      if (config.timeoutMs !== void 0 && config.timeoutMs > 0) {
-        timeout = setTimeout(
-          () => rejectWithError(`Speech synthesis timed out after ${config.timeoutMs} ms.`),
-          config.timeoutMs
-        );
-      }
-      synthesizer.speakSsmlAsync(ssml, cb, rejectWithError);
-    } catch (error) {
-      rejectWithError(error);
-    }
-  });
-}
-async function synthesizeSsmlChunks(chunks, config) {
-  const results = [];
-  const totalChunks = chunks.length;
-  const report = (event) => config.onProgress?.(event);
-  for (const [index, chunk] of chunks.entries()) {
-    const input = typeof chunk === "string" ? { ssml: chunk } : chunk;
-    report({
-      currentChunk: index,
-      totalChunks,
-      percent: totalChunks === 0 ? 100 : Math.round(index / totalChunks * 100),
-      chunkIndex: index,
-      originalTextRange: input.originalTextRange,
-      status: "pending",
-      durationMs: 0
-    });
-  }
-  for (const [index, chunk] of chunks.entries()) {
-    const input = typeof chunk === "string" ? { ssml: chunk } : chunk;
-    report({
-      currentChunk: index,
-      totalChunks,
-      percent: totalChunks === 0 ? 100 : Math.round(index / totalChunks * 100),
-      chunkIndex: index,
-      originalTextRange: input.originalTextRange,
-      status: "synthesizing",
-      durationMs: 0
-    });
-    const startedAt = Date.now();
-    try {
-      const result = await synthesizeSsml(input.ssml, {
-        ...config,
-        ...input.originalTextRange ? { sourceTextRange: input.originalTextRange } : {},
-        ...input.sourceNodePath ? { sourceNodePath: input.sourceNodePath } : {},
-        chunkIndex: index,
-        onProgress: void 0
-      });
-      results.push(result);
-      report({
-        currentChunk: index + 1,
-        totalChunks,
-        percent: totalChunks === 0 ? 100 : Math.round((index + 1) / totalChunks * 100),
-        chunkIndex: index,
-        originalTextRange: input.originalTextRange,
-        status: "success",
-        durationMs: Date.now() - startedAt
-      });
-    } catch (error) {
-      report({
-        currentChunk: index,
-        totalChunks,
-        percent: totalChunks === 0 ? 100 : Math.round(index / totalChunks * 100),
-        chunkIndex: index,
-        originalTextRange: input.originalTextRange,
-        status: "failed",
-        durationMs: Date.now() - startedAt,
-        error
-      });
-      throw error;
-    }
-  }
-  return mergeSynthesisResults(results, config.outputFormat ?? "audio-16khz-128kbitrate-mono-mp3");
-}
-function mergeSynthesisResults(results, format) {
-  const audioData = format ? new Uint8Array(
-    mergeAudioBuffers(
-      results.map((result) => result.audioData),
-      format
-    )
-  ) : new Uint8Array(results.reduce((total, result) => total + result.audioData.byteLength, 0));
-  if (!format) {
-    let offset = 0;
-    for (const result of results) {
-      audioData.set(new Uint8Array(result.audioData), offset);
-      offset += result.audioData.byteLength;
-    }
-  }
-  const boundaries = [];
-  const visemes = [];
-  const bookmarks = [];
-  let durationOffset = 0;
-  for (const result of results) {
-    const chunkBoundaries = result.boundaries && result.boundaries.length > 0 ? result.boundaries : result.wordBoundary ?? result.wordBoundaries ?? [];
-    for (const boundary of chunkBoundaries) {
-      const textRange = boundary.textRange ?? result.textRange;
-      const originalTextRange = boundary.originalTextRange ?? textRange;
-      const requestId = boundary.requestId ?? result.requestId;
-      boundaries.push({
-        ...boundary,
-        audioOffsetMs: boundary.audioOffsetMs + durationOffset,
-        chunkAudioOffsetMs: boundary.chunkAudioOffsetMs ?? boundary.audioOffsetMs,
-        ...boundary.chunkIndex === void 0 ? { chunkIndex: results.indexOf(result) } : {},
-        ...boundary.sourceNodePath ? { sourceNodePath: [...boundary.sourceNodePath] } : {},
-        ...originalTextRange ? { originalTextRange: { ...originalTextRange } } : {},
-        ...textRange ? { textRange: { ...textRange } } : {},
-        ...requestId ? { requestId } : {}
-      });
-    }
-    for (const viseme of result.visemes ?? []) {
-      const textRange = viseme.textRange ?? result.textRange;
-      const originalTextRange = viseme.originalTextRange ?? textRange;
-      const requestId = viseme.requestId ?? result.requestId;
-      visemes.push({
-        ...viseme,
-        audioOffsetMs: viseme.audioOffsetMs + durationOffset,
-        chunkAudioOffsetMs: viseme.chunkAudioOffsetMs ?? viseme.audioOffsetMs,
-        ...viseme.chunkIndex === void 0 ? { chunkIndex: results.indexOf(result) } : {},
-        ...viseme.sourceNodePath ? { sourceNodePath: [...viseme.sourceNodePath] } : {},
-        ...originalTextRange ? { originalTextRange: { ...originalTextRange } } : {},
-        ...textRange ? { textRange: { ...textRange } } : {},
-        ...requestId ? { requestId } : {}
-      });
-    }
-    for (const bookmark of result.bookmarks ?? []) {
-      const textRange = bookmark.textRange ?? result.textRange;
-      const originalTextRange = bookmark.originalTextRange ?? textRange;
-      const requestId = bookmark.requestId ?? result.requestId;
-      bookmarks.push({
-        ...bookmark,
-        audioOffsetMs: bookmark.audioOffsetMs + durationOffset,
-        chunkAudioOffsetMs: bookmark.chunkAudioOffsetMs ?? bookmark.audioOffsetMs,
-        ...bookmark.chunkIndex === void 0 ? { chunkIndex: results.indexOf(result) } : {},
-        ...bookmark.sourceNodePath ? { sourceNodePath: [...bookmark.sourceNodePath] } : {},
-        ...originalTextRange ? { originalTextRange: { ...originalTextRange } } : {},
-        ...textRange ? { textRange: { ...textRange } } : {},
-        ...requestId ? { requestId } : {}
-      });
-    }
-    durationOffset += Math.max(0, result.durationMs);
-  }
-  return {
-    audioData: audioData.buffer,
-    durationMs: durationOffset,
-    ...boundaries.length > 0 ? { boundaries, wordBoundary: boundaries, wordBoundaries: boundaries } : {},
-    ...visemes.length > 0 ? { visemes } : {},
-    ...bookmarks.length > 0 ? { bookmarks } : {},
-    ...results.length === 1 && results[0]?.requestId ? { requestId: results[0].requestId } : {},
-    ...results.length === 1 && results[0]?.textRange ? { textRange: { ...results[0].textRange } } : {}
-  };
-}
-async function synthesizeSpeech(ssml, config) {
-  return (await synthesizeSsml(ssml, config)).audioData;
-}
-
-// packages/ssml-core/dist/index.mjs
-var __typeError2 = (msg) => {
-  throw TypeError(msg);
-};
-var __accessCheck2 = (obj, member, msg) => member.has(obj) || __typeError2("Cannot " + msg);
-var __privateGet2 = (obj, member, getter) => (__accessCheck2(obj, member, "read from private field"), getter ? getter.call(obj) : member.get(obj));
-var __privateAdd2 = (obj, member, value) => member.has(obj) ? __typeError2("Cannot add the same private member more than once") : member instanceof WeakSet ? member.add(obj) : member.set(obj, value);
-var __privateSet2 = (obj, member, value, setter) => (__accessCheck2(obj, member, "write to private field"), setter ? setter.call(obj, value) : member.set(obj, value), value);
-var SYNTHESIS_NAMESPACE2 = "http://www.w3.org/2001/10/synthesis";
-var MSTTS_NAMESPACE2 = "http://www.w3.org/2001/mstts";
-var MAX_NESTING_DEPTH2 = 1e3;
-var SSML_TAGS2 = {
-  SPEAK: "speak",
-  VOICE: "voice",
-  PROSODY: "prosody",
-  BREAK: "break",
-  EXPRESS_AS: "express-as",
-  EXPRESS_AS_CAMEL: "expressAs",
-  MSTTS_EXPRESS_AS: "mstts:express-as",
-  SAY_AS: "say-as",
-  SAY_AS_CAMEL: "sayAs",
-  PHONEME: "phoneme",
-  EMPHASIS: "emphasis",
-  AUDIO: "audio",
-  SUB: "sub",
-  LANG: "lang",
-  MARK: "mark",
-  BOOKMARK: "bookmark",
-  LEXICON: "lexicon",
-  PARAGRAPH: "p",
-  SENTENCE: "s",
-  WORD: "w",
-  MSTTS_SILENCE: "mstts:silence",
-  SILENCE: "silence",
-  MSTTS_VISEME: "mstts:viseme",
-  VISEME: "viseme",
-  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_ATTRS2 = {
-  VERSION: "version",
-  XMLNS: "xmlns",
-  XML_LANG: "xml:lang",
-  LANG: "lang",
-  MSTTS_XMLNS: "xmlns:mstts",
-  NAME: "name",
-  VOICE: "voice",
-  SPEAKER: "speaker",
-  EFFECT: "effect",
-  RATE: "rate",
-  PITCH: "pitch",
-  VOLUME: "volume",
-  CONTOUR: "contour",
-  RANGE: "range",
-  TIME: "time",
-  STRENGTH: "strength",
-  STYLE: "style",
-  STYLE_DEGREE: "styledegree",
-  STYLE_DEGREE_CAMEL: "styleDegree",
-  STYLE_DEGREE_HYPHEN: "style-degree",
-  ROLE: "role",
-  INTERPRET_AS: "interpret-as",
-  FORMAT: "format",
-  DETAIL: "detail",
-  ALPHABET: "alphabet",
-  PH: "ph",
-  LEVEL: "level",
-  SRC: "src",
-  DESC: "desc",
-  CLIP_BEGIN: "clipBegin",
-  CLIP_END: "clipEnd",
-  SPEED: "speed",
-  REPEAT_COUNT: "repeatCount",
-  REPEAT_DURATION: "repeatDuration",
-  SOUND_LEVEL: "soundLevel",
-  ALIAS: "alias",
-  MARK: "mark",
-  URI: "uri",
-  ID: "id",
-  MODEL: "model",
-  PROFILE: "profile",
-  URL: "url",
-  SPEAKER_PROFILE_ID: "speakerProfileId",
-  TYPE: "type",
-  VALUE: "value",
-  FADE_IN: "fadein",
-  FADE_OUT: "fadeout"
-};
-var XML_ENTITIES2 = {
-  amp: "&",
-  apos: "'",
-  gt: ">",
-  lt: "<",
-  quot: '"'
-};
-function hasOwn2(object, property) {
-  return Object.getOwnPropertyDescriptor(object, property) !== void 0;
-}
-function setAttribute2(attributes, name, value) {
-  Object.defineProperty(attributes, name, {
-    configurable: true,
-    enumerable: true,
-    value,
-    writable: true
-  });
-}
-function decodeEntity2(entity) {
-  const namedValue = hasOwn2(XML_ENTITIES2, entity) ? XML_ENTITIES2[entity] : void 0;
-  if (namedValue !== void 0) {
-    return namedValue;
-  }
-  const isHexadecimal = entity.startsWith("#x") || entity.startsWith("#X");
-  const isDecimal = entity.startsWith("#");
-  if (!isHexadecimal && !isDecimal) {
-    throw new Error(`Unknown XML entity: &${entity};`);
-  }
-  const digits = entity.slice(isHexadecimal ? 2 : 1);
-  const codePoint = Number.parseInt(digits, isHexadecimal ? 16 : 10);
-  if (!digits || !Number.isInteger(codePoint) || codePoint < 0 || codePoint > 1114111 || codePoint >= 55296 && codePoint <= 57343 || codePoint < 32 && ![9, 10, 13].includes(codePoint)) {
-    throw new Error(`Invalid XML character reference: &${entity};`);
-  }
-  return String.fromCodePoint(codePoint);
-}
-function decodeXmlEntities2(value) {
-  let result = "";
-  let start = 0;
-  while (true) {
-    const ampersand = value.indexOf("&", start);
-    if (ampersand === -1) {
-      return result + value.slice(start);
-    }
-    result += value.slice(start, ampersand);
-    const semicolon = value.indexOf(";", ampersand + 1);
-    if (semicolon === -1) {
-      throw new Error("Unterminated XML entity reference");
-    }
-    result += decodeEntity2(value.slice(ampersand + 1, semicolon));
-    start = semicolon + 1;
-  }
-}
-function isXmlNameStart2(value) {
-  return value !== void 0 && /[A-Za-z_]/.test(value);
-}
-function isXmlNameCharacter2(value) {
-  return value !== void 0 && /[A-Za-z0-9_.:-]/.test(value);
-}
-function isXmlWhitespace2(value) {
-  return value === " " || value === "	" || value === "\r" || value === "\n";
-}
-function removeStandardNamespaceAttributes2(attributes) {
-  if (attributes[SSML_ATTRS2.XMLNS] === SYNTHESIS_NAMESPACE2) {
-    delete attributes[SSML_ATTRS2.XMLNS];
-  }
-  if (attributes[SSML_ATTRS2.MSTTS_XMLNS] === MSTTS_NAMESPACE2) {
-    delete attributes[SSML_ATTRS2.MSTTS_XMLNS];
-  }
-}
-var _index2;
-var XmlParser2 = class {
-  constructor(source) {
-    __privateAdd2(this, _index2, 0);
-    this.source = source;
-  }
-  parse() {
-    if (this.source.charCodeAt(0) === 65279) {
-      __privateSet2(this, _index2, __privateGet2(this, _index2) + 1);
-    }
-    this.skipMisc();
-    if (__privateGet2(this, _index2) >= this.source.length) {
-      this.fail("SSML input is empty");
-    }
-    if (this.source[__privateGet2(this, _index2)] !== "<") {
-      this.fail("SSML input must start with an XML element");
-    }
-    const root = this.parseElement(0);
-    this.skipMisc();
-    if (__privateGet2(this, _index2) !== this.source.length) {
-      this.fail("Unexpected content after the root XML element");
-    }
-    return root;
+  parse() {
+    if (this.source.charCodeAt(0) === 65279) {
+      __privateSet2(this, _index2, __privateGet2(this, _index2) + 1);
+    }
+    this.skipMisc();
+    if (__privateGet2(this, _index2) >= this.source.length) {
+      this.fail("SSML input is empty");
+    }
+    if (this.source[__privateGet2(this, _index2)] !== "<") {
+      this.fail("SSML input must start with an XML element");
+    }
+    const root = this.parseElement(0);
+    this.skipMisc();
+    if (__privateGet2(this, _index2) !== this.source.length) {
+      this.fail("Unexpected content after the root XML element");
+    }
+    return root;
   }
   parseElement(depth) {
     if (depth > MAX_NESTING_DEPTH2) {
@@ -3702,6 +3460,130 @@ function parseSsml2(xmlString) {
   }
   return document;
 }
+function decodeXmlText2(value) {
+  return value.replace(/&(?:amp|apos|gt|lt|quot);|&#(?:x[\da-f]+|\d+);/gi, (entity) => {
+    if (entity === "&amp;") return "&";
+    if (entity === "&apos;") return "'";
+    if (entity === "&gt;") return ">";
+    if (entity === "&lt;") return "<";
+    if (entity === "&quot;") return '"';
+    const hexadecimal = entity.toLowerCase().startsWith("&#x");
+    const digits = entity.slice(hexadecimal ? 3 : 2, -1);
+    return String.fromCodePoint(Number.parseInt(digits, hexadecimal ? 16 : 10));
+  });
+}
+function decodeXmlAttribute2(value) {
+  return decodeXmlText2(value);
+}
+function findTagEnd3(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 readTagName2(tag) {
+  const match = /^<\s*([A-Za-z_][A-Za-z0-9_.:-]*)/.exec(tag);
+  return match?.[1];
+}
+function readTagAttributes2(tag, name) {
+  const attributes = {};
+  const nameStart = tag.indexOf(name);
+  const attributeSource = tag.slice(nameStart + name.length, tag.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[match[1].toLowerCase()] = decodeXmlAttribute2(match[3]);
+  }
+  return attributes;
+}
+function collectSourceMap2(source) {
+  const segments = [];
+  const markers = [];
+  const elements = [];
+  let textOffset = 0;
+  let index = 0;
+  const textParts = [];
+  const addText = (value) => {
+    if (!value) return;
+    const parent = elements[elements.length - 1];
+    if (parent) parent.nextChildIndex += 1;
+    const sourceNodePath = parent?.path ?? ["speak"];
+    const start = textOffset;
+    textOffset += value.length;
+    textParts.push(value);
+    segments.push({ text: value, range: { start, end: textOffset }, sourceNodePath: [...sourceNodePath] });
+  };
+  while (index < source.length) {
+    if (source[index] !== "<") {
+      const end2 = source.indexOf("<", index);
+      const textEnd = end2 === -1 ? source.length : end2;
+      addText(decodeXmlText2(source.slice(index, textEnd)));
+      index = textEnd;
+      continue;
+    }
+    if (source.startsWith("<!--", index)) {
+      const end2 = source.indexOf("-->", index + 4);
+      index = end2 === -1 ? source.length : end2 + 3;
+      continue;
+    }
+    if (source.startsWith("<![CDATA[", index)) {
+      const contentStart = index + 9;
+      const end2 = source.indexOf("]]>", contentStart);
+      const contentEnd = end2 === -1 ? source.length : end2;
+      addText(source.slice(contentStart, contentEnd));
+      index = end2 === -1 ? source.length : end2 + 3;
+      continue;
+    }
+    if (source.startsWith("<?", index)) {
+      const end2 = source.indexOf("?>", index + 2);
+      index = end2 === -1 ? source.length : end2 + 2;
+      continue;
+    }
+    const end = findTagEnd3(source, index + 1);
+    const rawTag = source.slice(index, end + 1);
+    if (rawTag.startsWith("</")) {
+      elements.pop();
+      index = end + 1;
+      continue;
+    }
+    const name = readTagName2(rawTag);
+    if (!name) {
+      index = end + 1;
+      continue;
+    }
+    const parent = elements[elements.length - 1];
+    const childIndex = parent?.nextChildIndex ?? 0;
+    if (parent) parent.nextChildIndex += 1;
+    const path = parent ? [...parent.path, `${name}[${childIndex}]`] : [name];
+    const attributes = readTagAttributes2(rawTag, name);
+    const normalizedName = name.toLowerCase();
+    if (normalizedName === "mark" || normalizedName === "bookmark") {
+      const markerName = attributes[normalizedName === "mark" ? "name" : "mark"];
+      if (markerName) {
+        markers.push({
+          kind: normalizedName,
+          name: markerName,
+          originalTextRange: { start: textOffset, end: textOffset },
+          sourceNodePath: [...path]
+        });
+      }
+    }
+    if (!/\/\s*>$/.test(rawTag)) elements.push({ name, path, nextChildIndex: 0 });
+    index = end + 1;
+  }
+  return { text: textParts.join(""), segments, markers };
+}
+function getSsmlSourceMap2(ssml) {
+  parseSsml2(ssml);
+  return collectSourceMap2(ssml);
+}
 var AZURE_VOICE_DEFINITIONS2 = [
   { name: "de-DE-ConradNeural", locale: "de-DE", styles: ["cheerful", "sad"] },
   { name: "de-DE-KatjaNeural", locale: "de-DE", styles: ["cheerful", "sad"] },
@@ -3829,34 +3711,51 @@ function createAzureUrlValidatorRunner2(validator, options = {}) {
   const inFlight = /* @__PURE__ */ new Map();
   const waiters = [];
   let active = 0;
-  const acquire = async () => {
+  const configuredSignal = options.signal ?? new AbortController().signal;
+  const acquire = async (signal) => {
+    if (signal.aborted) throw new Error("URL validation was aborted.");
     if (active < concurrency) {
       active += 1;
       return;
     }
-    await new Promise((resolve) => waiters.push(resolve));
+    await new Promise((resolve, reject) => {
+      let waiter;
+      const abortHandler = () => {
+        const index = waiters.indexOf(waiter);
+        if (index >= 0) waiters.splice(index, 1);
+        signal.removeEventListener("abort", abortHandler);
+        reject(new Error("URL validation was aborted."));
+      };
+      signal.addEventListener("abort", abortHandler, { once: true });
+      waiter = () => {
+        signal.removeEventListener("abort", abortHandler);
+        resolve();
+      };
+      waiters.push(waiter);
+    });
     active += 1;
   };
   const release = () => {
     active -= 1;
     waiters.shift()?.();
   };
-  const check = async (url, context) => {
-    if (options.signal?.aborted) throw new Error("URL validation was aborted.");
-    const cached = cache.get(url);
+  const check = async (url, context, signal = configuredSignal) => {
+    if (signal.aborted) throw new Error("URL validation was aborted.");
+    const key = `${context.tag}:${context.attribute}:${url}`;
+    const cached = cache.get(key);
     if (cached !== void 0) return cached;
-    const existing = inFlight.get(url);
+    const existing = inFlight.get(key);
     if (existing) return existing;
     const promise = (async () => {
-      await acquire();
+      await acquire(signal);
       try {
-        if (options.signal?.aborted) throw new Error("URL validation was aborted.");
-        const validation = Promise.resolve(validator(url, context));
+        if (signal.aborted) throw new Error("URL validation was aborted.");
+        const validation = Promise.resolve(validator(url, context, signal));
         let timer;
         let abortHandler;
         const cancellation = new Promise((_resolve, reject) => {
           abortHandler = () => reject(new Error("URL validation was aborted."));
-          options.signal?.addEventListener("abort", abortHandler, { once: true });
+          signal.addEventListener("abort", abortHandler, { once: true });
           if (options.timeoutMs !== void 0 && options.timeoutMs > 0) {
             timer = setTimeout(
               () => reject(new Error(`URL validation timed out after ${options.timeoutMs} ms.`)),
@@ -3866,24 +3765,24 @@ function createAzureUrlValidatorRunner2(validator, options = {}) {
         });
         try {
           const result = await (timer || abortHandler ? Promise.race([validation, cancellation]) : validation);
-          cache.set(url, result);
+          cache.set(key, result);
           return result;
         } finally {
           if (timer) clearTimeout(timer);
-          if (abortHandler) options.signal?.removeEventListener("abort", abortHandler);
+          if (abortHandler) signal.removeEventListener("abort", abortHandler);
         }
       } finally {
         release();
       }
     })();
-    inFlight.set(url, promise);
+    inFlight.set(key, promise);
     try {
       return await promise;
     } finally {
-      inFlight.delete(url);
+      inFlight.delete(key);
     }
   };
-  return (url, context) => check(url, context);
+  return (url, context, signal) => check(url, context, signal ?? configuredSignal);
 }
 var ALLOWED_BREAK_STRENGTHS2 = /* @__PURE__ */ new Set(["none", "x-weak", "weak", "medium", "strong", "x-strong"]);
 var ALLOWED_SAY_AS2 = /* @__PURE__ */ new Set([
@@ -3996,6 +3895,7 @@ function tokenizeElements2(source) {
     if (parent) parent.childElementCount += 1;
     const parentVoiceName = [...openElements].reverse().find((element) => element.voiceName)?.voiceName;
     const tokenName = nameMatch[1];
+    const path = parent ? [...parent.path, `${tokenName}[${childElementIndex ?? 0}]`] : [tokenName];
     const tokenVoiceName = tokenName.toLowerCase() === "voice" ? attributes.get("name") : tokenName.toLowerCase() === "mstts:turn" ? attributes.get("voice") ?? parentVoiceName : parentVoiceName;
     tokens.push({
       attributes,
@@ -4006,12 +3906,14 @@ function tokenizeElements2(source) {
       parentName: parent?.name,
       parentVoiceName,
       selfClosing,
-      start
+      start,
+      path
     });
     if (!selfClosing) {
       openElements.push({
         childElementCount: 0,
         name: tokenName,
+        path,
         voiceName: tokenVoiceName
       });
     }
@@ -4024,13 +3926,14 @@ function location2(source, offset) {
   const line = before.split("\n").length;
   return { line, column: before.length - (before.lastIndexOf("\n") + 1) + 1 };
 }
-function addDiagnostic2(diagnostics, source, offset, message, severity = "error", code) {
+function addDiagnostic2(diagnostics, source, offset, message, severity = "error", code, metadata = {}) {
   diagnostics.push({
     ...location2(source, offset),
     message,
     severity,
     source: "ssml-static-validator",
-    ...code ? { code } : {}
+    ...code ? { code } : {},
+    ...metadata
   });
 }
 function isSupportedProsodyRate2(value) {
@@ -4331,279 +4234,1184 @@ function validateElement2(token, source, diagnostics, voiceName, options, voiceC
     if (!value || !isValidAzureAudioDuration2(value))
       addDiagnostic2(
         diagnostics,
-        source,
+        source,
+        token.start,
+        '<mstts:audioduration> requires a positive duration such as "10s", "5000ms", or "00:00:10".'
+      );
+    if (!token.selfClosing)
+      addDiagnostic2(diagnostics, source, token.start, "<mstts:audioduration> must be self-closing.");
+  }
+  if (name === "mstts:viseme") {
+    const type = attr2(token, "type");
+    if (!type || !ALLOWED_VISEME_TYPES2.has(type))
+      addDiagnostic2(diagnostics, source, token.start, '<mstts:viseme> requires a supported "type" attribute.');
+  }
+  if (name === "audio") {
+    validateAudioSource2(token, source, diagnostics, options, "audio");
+  }
+  if (name === "mstts:turn") {
+    if (!attr2(token, "voice")?.trim() && !attr2(token, "speaker")?.trim())
+      addDiagnostic2(
+        diagnostics,
+        source,
+        token.start,
+        '<mstts:turn> requires a non-empty "voice" or "speaker" attribute.'
+      );
+    if (token.parentName?.toLowerCase() !== "mstts:dialog")
+      addDiagnostic2(diagnostics, source, token.start, "<mstts:turn> is only allowed directly inside <mstts:dialog>.");
+  }
+  if (name === "mstts:backgroundaudio") {
+    validateAudioSource2(token, source, diagnostics, options, "mstts:backgroundaudio");
+    const volume = attr2(token, "volume");
+    if (volume !== void 0 && (!/^\d+(?:\.\d+)?$/.test(volume.trim()) || Number(volume) > 100))
+      addDiagnostic2(diagnostics, source, token.start, `Unsupported <mstts:backgroundaudio volume> value "${volume}".`);
+    for (const [attribute, value] of [
+      ["fadein", attr2(token, "fadein")],
+      ["fadeout", attr2(token, "fadeout")]
+    ]) {
+      if (value !== void 0 && !isValidAzureBackgroundAudioDuration2(value))
+        addDiagnostic2(
+          diagnostics,
+          source,
+          token.start,
+          `<mstts:backgroundaudio ${attribute}> must be between 0 and 10000 milliseconds, for example "500ms" or "10s".`
+        );
+    }
+    if (token.parentName?.toLowerCase() !== "speak" || token.childElementIndex !== 0)
+      addDiagnostic2(
+        diagnostics,
+        source,
+        token.start,
+        "<mstts:backgroundaudio> must be the first element directly under <speak>."
+      );
+    if (!token.selfClosing)
+      addDiagnostic2(diagnostics, source, token.start, "<mstts:backgroundaudio> must be self-closing.");
+  }
+}
+function validateAzureSsmlStatic2(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)
+    addDiagnostic2(diagnostics, ssml, maxLength, `SSML exceeds the maximum length of ${maxLength} characters.`);
+  if (options.maxXmlDepth !== void 0 && (!Number.isInteger(options.maxXmlDepth) || options.maxXmlDepth <= 0)) {
+    addDiagnostic2(diagnostics, ssml, 0, "maxXmlDepth must be a positive integer.");
+  }
+  try {
+    parseSsml2(ssml);
+  } catch (error) {
+    const message = error instanceof Error ? error.message.replace(/ at position \d+$/, "") : String(error);
+    const match = / at position (\d+)$/.exec(error instanceof Error ? error.message : "");
+    addDiagnostic2(diagnostics, ssml, match ? Number(match[1]) : 0, message);
+    return diagnostics;
+  }
+  const tokens = tokenizeElements2(ssml);
+  if (options.maxXmlDepth !== void 0) {
+    for (const token of tokens) {
+      if (token.depth > options.maxXmlDepth) {
+        addDiagnostic2(
+          diagnostics,
+          ssml,
+          token.start,
+          `XML nesting depth ${token.depth} exceeds the configured maximum of ${options.maxXmlDepth}.`
+        );
+      }
+    }
+  }
+  const speak = tokens.find((token) => token.name.toLowerCase() === "speak");
+  const voices = tokens.filter((token) => token.name.toLowerCase() === "voice");
+  const backgroundAudioTokens = tokens.filter((token) => token.name.toLowerCase() === "mstts:backgroundaudio");
+  for (const [index, token] of backgroundAudioTokens.entries()) {
+    if (index > 0)
+      addDiagnostic2(
+        diagnostics,
+        ssml,
+        token.start,
+        "An SSML document can contain at most one <mstts:backgroundaudio> element."
+      );
+  }
+  if (!speak || voices.length === 0)
+    addDiagnostic2(
+      diagnostics,
+      ssml,
+      speak?.start ?? 0,
+      "Azure SSML requires at least one <voice> element under <speak>."
+    );
+  const voiceName = voices[0] ? attr2(voices[0], "name") : void 0;
+  const voiceCatalog = normalizeVoiceCatalog2(options);
+  const normalizeLanguage = createLanguageNormalizer2(options);
+  const policySeverity = diagnosticSeverity2(options.unknownVoicePolicy ?? "warn");
+  const voicesToValidate = options.validateNestedVoices === false ? voices.slice(0, 1) : voices;
+  for (const token of voicesToValidate) {
+    const name = attr2(token, "name")?.trim();
+    const language = attr2(token, "xml:lang")?.trim() || (speak ? attr2(speak, "xml:lang")?.trim() : void 0);
+    const definition = name ? voiceCatalog.get(name.toLowerCase()) : void 0;
+    if (name && definition?.status === "preview")
+      addDiagnostic2(
+        diagnostics,
+        ssml,
+        token.start,
+        `Voice "${name}" is an Azure Speech preview voice and may change or require preview access.`,
+        "warning",
+        "azure-preview-voice"
+      );
+    if (name && definition?.status === "deprecated")
+      addDiagnostic2(
+        diagnostics,
+        ssml,
+        token.start,
+        `Voice "${name}" is deprecated by Azure Speech; migrate to a supported voice.`,
+        "info",
+        "azure-deprecated-voice"
+      );
+    if (name && !definition && policySeverity)
+      addDiagnostic2(
+        diagnostics,
+        ssml,
+        token.start,
+        `Unknown voice "${name}" is not registered in the voice catalog.`,
+        policySeverity,
+        "azure-unknown-voice"
+      );
+    if (name && language && definitionMatchesLanguage2(definition, name, language, normalizeLanguage) === false)
+      addDiagnostic2(
+        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" ? attr2(token, "name")?.trim() : tokenName === "mstts:turn" ? attr2(token, "voice")?.trim() || token.parentVoiceName : options.validateNestedVoices === false ? voiceName : token.parentVoiceName;
+    const tokenDiagnosticStart = diagnostics.length;
+    validateElement2(token, ssml, diagnostics, tokenVoiceName, options, voiceCatalog);
+    const definition = tokenVoiceName ? voiceCatalog.get(tokenVoiceName.toLowerCase()) : void 0;
+    validateVoiceFeatureMatrix2(token, ssml, diagnostics, tokenVoiceName, definition);
+    annotateTokenDiagnostics2(diagnostics, tokenDiagnosticStart, token, options, tokenVoiceName);
+    if (tokenName === "voice" && options.model && definition?.models && !definition.models.some((model) => model.toLowerCase() === options.model?.toLowerCase())) {
+      addDiagnostic2(
+        diagnostics,
+        ssml,
         token.start,
-        '<mstts:audioduration> requires a positive duration such as "10s", "5000ms", or "00:00:10".'
+        `Voice "${tokenVoiceName}" does not support model "${options.model}" according to the configured feature matrix.`,
+        "error",
+        "azure-unsupported-model-for-voice"
       );
-    if (!token.selfClosing)
-      addDiagnostic2(diagnostics, source, token.start, "<mstts:audioduration> must be self-closing.");
-  }
-  if (name === "mstts:viseme") {
-    const type = attr2(token, "type");
-    if (!type || !ALLOWED_VISEME_TYPES2.has(type))
-      addDiagnostic2(diagnostics, source, token.start, '<mstts:viseme> requires a supported "type" attribute.');
+    }
   }
-  if (name === "audio") {
-    validateAudioSource2(token, source, diagnostics, options, "audio");
+  return diagnostics;
+}
+function urlAttributes2(token) {
+  const tag = canonicalTagName2(token.name);
+  const attributes = tag === "audio" || tag === "mstts:backgroundaudio" ? ["src"] : tag === "lexicon" ? ["uri"] : tag === "mstts:voiceconversion" ? ["url"] : [];
+  return attributes.flatMap((attribute) => {
+    const value = attr2(token, attribute);
+    return value === void 0 ? [] : [{ attribute, value }];
+  });
+}
+function annotateTokenDiagnostics2(diagnostics, startIndex, token, options, voiceName) {
+  const attributes = [...token.attributes.keys()];
+  for (const diagnostic of diagnostics.slice(startIndex)) {
+    const attributeName = attributes.find(
+      (attribute) => new RegExp(`(?:<[^> ]+\\s+|")${attribute}(?:"|>|\\s)`, "i").test(diagnostic.message)
+    );
+    const nodePath = options.sourceNodePath ? [...options.sourceNodePath] : [...token.path];
+    Object.assign(diagnostic, {
+      range: { start: token.start, end: token.end + 1 },
+      tagName: token.name,
+      ...attributeName ? { attributeName } : {},
+      ...voiceName ? { voiceName } : {},
+      ...options.chunkIndex !== void 0 ? { chunkIndex: options.chunkIndex } : {},
+      nodePath,
+      targetNodePath: [...token.path]
+    });
   }
-  if (name === "mstts:turn") {
-    if (!attr2(token, "voice")?.trim() && !attr2(token, "speaker")?.trim())
-      addDiagnostic2(
-        diagnostics,
-        source,
-        token.start,
-        '<mstts:turn> requires a non-empty "voice" or "speaker" attribute.'
-      );
-    if (token.parentName?.toLowerCase() !== "mstts:dialog")
-      addDiagnostic2(diagnostics, source, token.start, "<mstts:turn> is only allowed directly inside <mstts:dialog>.");
+}
+function validateAzureSsml2(ssml, options = {}) {
+  const diagnostics = validateAzureSsmlStatic2(ssml, options);
+  const validator = options.urlValidator ?? options.customUrlValidator;
+  if (!validator || typeof ssml !== "string") return diagnostics;
+  const runnerOptions = options.urlValidation ?? {};
+  const boundedValidator = options.urlValidatorRunner ?? createAzureUrlValidatorRunner2(validator, {
+    ...runnerOptions,
+    ...options.urlValidatorConcurrency !== void 0 ? { concurrency: options.urlValidatorConcurrency } : {},
+    ...options.urlValidatorTimeoutMs !== void 0 ? { timeoutMs: options.urlValidatorTimeoutMs } : {},
+    ...options.urlValidatorSignal ? { signal: options.urlValidatorSignal } : {},
+    ...options.urlValidatorCache ? { cache: options.urlValidatorCache } : {}
+  });
+  const validationSignal = options.urlValidatorSignal ?? options.urlValidation?.signal ?? new AbortController().signal;
+  let tokens;
+  try {
+    tokens = tokenizeElements2(ssml);
+  } catch {
+    return diagnostics;
   }
-  if (name === "mstts:backgroundaudio") {
-    validateAudioSource2(token, source, diagnostics, options, "mstts:backgroundaudio");
-    const volume = attr2(token, "volume");
-    if (volume !== void 0 && (!/^\d+(?:\.\d+)?$/.test(volume.trim()) || Number(volume) > 100))
-      addDiagnostic2(diagnostics, source, token.start, `Unsupported <mstts:backgroundaudio volume> value "${volume}".`);
-    for (const [attribute, value] of [
-      ["fadein", attr2(token, "fadein")],
-      ["fadeout", attr2(token, "fadeout")]
-    ]) {
-      if (value !== void 0 && !isValidAzureBackgroundAudioDuration2(value))
+  const checks = tokens.flatMap(
+    (token) => urlAttributes2(token).map(async ({ attribute, value }) => {
+      try {
+        const result = await boundedValidator(
+          value,
+          { tag: token.name, attribute, ...options.sourceNodePath ? { sourceNodePath: options.sourceNodePath } : {} },
+          validationSignal
+        );
+        const valid = typeof result === "boolean" ? result : result.valid;
+        if (!valid) {
+          const reason = typeof result === "boolean" ? void 0 : result.reason;
+          const diagnosticStart = diagnostics.length;
+          addDiagnostic2(
+            diagnostics,
+            ssml,
+            token.start,
+            `<${token.name} ${attribute}> was rejected by the custom URL validator${reason ? `: ${reason}` : "."}`
+          );
+          annotateTokenDiagnostics2(diagnostics, diagnosticStart, token, options, token.parentVoiceName);
+        }
+      } catch (error) {
+        const reason = error instanceof Error ? error.message : String(error);
+        const diagnosticStart = diagnostics.length;
         addDiagnostic2(
           diagnostics,
-          source,
+          ssml,
           token.start,
-          `<mstts:backgroundaudio ${attribute}> must be between 0 and 10000 milliseconds, for example "500ms" or "10s".`
+          `<${token.name} ${attribute}> could not be validated by the custom URL validator: ${reason}`
         );
+        annotateTokenDiagnostics2(diagnostics, diagnosticStart, token, options, token.parentVoiceName);
+      }
+    })
+  );
+  return Promise.all(checks).then(() => diagnostics);
+}
+var AZURE_VOICE_CATALOG_METADATA2 = {
+  apiVersion: "2025-10-01",
+  generatedAt: "2026-08-28T00:00:00.000Z",
+  regions: [],
+  voiceCount: AZURE_VOICE_DEFINITIONS2.length
+};
+
+// packages/azure-tts-client/src/outputFormats.ts
+var SpeechSDK = __toESM(require("microsoft-cognitiveservices-speech-sdk"));
+var DEFAULT_OUTPUT_FORMAT = "audio-16khz-128kbitrate-mono-mp3";
+var OUTPUT_FORMATS = {
+  "raw-8khz-8bit-mono-mulaw": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz8BitMonoMULaw,
+  "riff-16khz-16kbps-mono-siren": SpeechSDK.SpeechSynthesisOutputFormat.Riff16Khz16KbpsMonoSiren,
+  "audio-16khz-16kbps-mono-siren": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz16KbpsMonoSiren,
+  "audio-16khz-32kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz32KBitRateMonoMp3,
+  "audio-16khz-128kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz128KBitRateMonoMp3,
+  "audio-16khz-64kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz64KBitRateMonoMp3,
+  "audio-24khz-48kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz48KBitRateMonoMp3,
+  "audio-24khz-96kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz96KBitRateMonoMp3,
+  "audio-24khz-160kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz160KBitRateMonoMp3,
+  "raw-16khz-16bit-mono-truesilk": SpeechSDK.SpeechSynthesisOutputFormat.Raw16Khz16BitMonoTrueSilk,
+  "riff-16khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Riff16Khz16BitMonoPcm,
+  "riff-8khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz16BitMonoPcm,
+  "riff-24khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Riff24Khz16BitMonoPcm,
+  "riff-8khz-8bit-mono-mulaw": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz8BitMonoMULaw,
+  "raw-16khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Raw16Khz16BitMonoPcm,
+  "raw-24khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Raw24Khz16BitMonoPcm,
+  "raw-8khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz16BitMonoPcm,
+  "ogg-16khz-16bit-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Ogg16Khz16BitMonoOpus,
+  "ogg-24khz-16bit-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Ogg24Khz16BitMonoOpus,
+  "raw-48khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Raw48Khz16BitMonoPcm,
+  "riff-48khz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Riff48Khz16BitMonoPcm,
+  "audio-48khz-96kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio48Khz96KBitRateMonoMp3,
+  "audio-48khz-192kbitrate-mono-mp3": SpeechSDK.SpeechSynthesisOutputFormat.Audio48Khz192KBitRateMonoMp3,
+  "ogg-48khz-16bit-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Ogg48Khz16BitMonoOpus,
+  "webm-16khz-16bit-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Webm16Khz16BitMonoOpus,
+  "webm-24khz-16bit-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Webm24Khz16BitMonoOpus,
+  "webm-24khz-16bit-24kbps-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Webm24Khz16Bit24KbpsMonoOpus,
+  "raw-24khz-16bit-mono-truesilk": SpeechSDK.SpeechSynthesisOutputFormat.Raw24Khz16BitMonoTrueSilk,
+  "raw-8khz-8bit-mono-alaw": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz8BitMonoALaw,
+  "riff-8khz-8bit-mono-alaw": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz8BitMonoALaw,
+  "audio-16khz-16bit-32kbps-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz16Bit32KbpsMonoOpus,
+  "audio-24khz-16bit-48kbps-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz16Bit48KbpsMonoOpus,
+  "audio-24khz-16bit-24kbps-mono-opus": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz16Bit24KbpsMonoOpus,
+  "raw-22050hz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Raw22050Hz16BitMonoPcm,
+  "riff-22050hz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Riff22050Hz16BitMonoPcm,
+  "raw-44100hz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Raw44100Hz16BitMonoPcm,
+  "riff-44100hz-16bit-mono-pcm": SpeechSDK.SpeechSynthesisOutputFormat.Riff44100Hz16BitMonoPcm,
+  "amr-wb-16000hz": SpeechSDK.SpeechSynthesisOutputFormat.AmrWb16000Hz,
+  "g722-16khz-64kbps": SpeechSDK.SpeechSynthesisOutputFormat.G72216Khz64Kbps
+};
+function resolveMimeType(outputFormat) {
+  if (/(?:wav|wave|riff)/i.test(outputFormat)) return "audio/wav";
+  if (/(?:mp3|mpeg)/i.test(outputFormat)) return "audio/mpeg";
+  if (/ogg/i.test(outputFormat)) return "audio/ogg";
+  if (/webm/i.test(outputFormat)) return "audio/webm";
+  if (/raw/i.test(outputFormat)) return "audio/L16";
+  return "application/octet-stream";
+}
+function resolveOutputFormat(outputFormat) {
+  const resolvedFormat = OUTPUT_FORMATS[outputFormat];
+  if (resolvedFormat === void 0) {
+    throw new Error(`Unsupported Azure Speech output format: ${outputFormat}`);
+  }
+  return resolvedFormat;
+}
+
+// packages/azure-tts-client/src/speechConfig.ts
+var import_microsoft_cognitiveservices_speech_sdk = require("microsoft-cognitiveservices-speech-sdk");
+function resolveEndpoint(config) {
+  const endpoint = config.endpoint?.trim() || "https://{region}.tts.speech.microsoft.com/cognitiveservices/v1";
+  return endpoint.replace(/\{region\}/g, encodeURIComponent(config.region));
+}
+function createSpeechConfig(config) {
+  const { outputFormat = DEFAULT_OUTPUT_FORMAT, subscriptionKey } = config;
+  const endpoint = new URL(resolveEndpoint(config));
+  const speechConfig = import_microsoft_cognitiveservices_speech_sdk.SpeechConfig.fromEndpoint(endpoint, subscriptionKey);
+  speechConfig.speechSynthesisOutputFormat = resolveOutputFormat(outputFormat);
+  return speechConfig;
+}
+
+// packages/azure-tts-client/src/synthesis.ts
+function ascii(bytes, offset, value) {
+  return [...value].every((character, index) => bytes[offset + index] === character.charCodeAt(0));
+}
+function readUint32(bytes, offset) {
+  return new DataView(bytes.buffer, bytes.byteOffset, bytes.byteLength).getUint32(offset, true);
+}
+function parseWav(buffer) {
+  const bytes = new Uint8Array(buffer);
+  if (bytes.byteLength < 12 || !ascii(bytes, 0, "RIFF") || !ascii(bytes, 8, "WAVE")) {
+    throw new Error("Invalid WAV/RIFF audio buffer.");
+  }
+  const chunks = [];
+  const dataParts = [];
+  let format;
+  let offset = 12;
+  while (offset < bytes.byteLength) {
+    if (offset + 8 > bytes.byteLength) throw new Error("Invalid WAV chunk header.");
+    const id = String.fromCharCode(...bytes.slice(offset, offset + 4));
+    const size = readUint32(bytes, offset + 4);
+    const dataStart = offset + 8;
+    const dataEnd = dataStart + size;
+    if (dataEnd > bytes.byteLength) throw new Error(`WAV chunk "${id}" exceeds the audio buffer.`);
+    const data2 = bytes.slice(dataStart, dataEnd);
+    chunks.push({ id, data: data2 });
+    if (id === "fmt ") format ?? (format = data2);
+    if (id === "data") dataParts.push(data2);
+    offset = dataEnd + (size & 1);
+    if (offset > bytes.byteLength) throw new Error("Invalid WAV chunk padding.");
+  }
+  if (!format || dataParts.length === 0) throw new Error("WAV audio must contain fmt and data chunks.");
+  const dataLength = dataParts.reduce((total, part) => total + part.byteLength, 0);
+  const data = new Uint8Array(dataLength);
+  let dataOffset = 0;
+  for (const part of dataParts) {
+    data.set(part, dataOffset);
+    dataOffset += part.byteLength;
+  }
+  return { chunks, data, format };
+}
+function formatNumber(format, pattern, fallback) {
+  const match = pattern.exec(format);
+  return match?.[1] ? Number(match[1]) : fallback;
+}
+function formatChannels(format, fallback) {
+  if (/stereo|2ch|dual/i.test(format)) return 2;
+  if (/mono|1ch/i.test(format)) return 1;
+  return fallback;
+}
+function formatAudioSpecification(format) {
+  const sampleRate = formatNumber(format, /(\d+)(?:khz|kHz|hz|Hz)/, 0);
+  const channels = formatChannels(format, 0);
+  const bitrateMatch = /(\d+)(?:kbitrate|kbps|kbits?)/i.exec(format);
+  const bitrate = bitrateMatch?.[1] ? Number(bitrateMatch[1]) * 1e3 : void 0;
+  const codec = /mp3|mpeg/i.test(format) ? "mp3" : /opus/i.test(format) ? "opus" : /silk/i.test(format) ? "silk" : /pcm|mulaw|alaw|siren/i.test(format) ? "pcm" : "unknown";
+  return {
+    format,
+    mimeType: resolveMimeType(format),
+    codec,
+    sampleRate,
+    channels,
+    ...bitrate ? { bitrate } : {},
+    isCompressed: codec === "mp3" || codec === "opus" || codec === "silk"
+  };
+}
+function parseMp3Specification(buffer, format) {
+  const bytes = stripMp3Tags(buffer);
+  const bitrates = [
+    [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+    [0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 0],
+    [0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, 0]
+  ];
+  const sampleRates = [
+    [44100, 48e3, 32e3],
+    [22050, 24e3, 16e3],
+    [11025, 12e3, 8e3]
+  ];
+  for (let index = 0; index + 4 <= bytes.length; index += 1) {
+    if (bytes[index] !== 255 || (bytes[index + 1] ?? 0) < 224) continue;
+    const header = bytes[index + 1] ?? 0;
+    const versionBits = header >> 3 & 3;
+    const layer = header >> 1 & 3;
+    const bitrateIndex = (bytes[index + 2] ?? 0) >> 4;
+    const sampleIndex = (bytes[index + 2] ?? 0) >> 2 & 3;
+    if (versionBits === 1 || layer !== 1 || bitrateIndex === 0 || bitrateIndex === 15 || sampleIndex === 3) continue;
+    const versionIndex = versionBits === 3 ? 0 : versionBits === 2 ? 1 : 2;
+    const bitrateTable = versionBits === 3 ? bitrates[1] : bitrates[2];
+    const sampleRate = sampleRates[versionIndex]?.[sampleIndex] ?? 0;
+    const bitrateKbps = bitrateTable?.[bitrateIndex] ?? 0;
+    if (!sampleRate || !bitrateKbps) continue;
+    return {
+      format,
+      mimeType: "audio/mpeg",
+      codec: "mp3",
+      sampleRate,
+      channels: (bytes[index + 3] ?? 0) >> 6 === 3 ? 1 : 2,
+      bitrate: bitrateKbps * 1e3,
+      isCompressed: true
+    };
+  }
+  return void 0;
+}
+function inspectAudioSpecification(buffer, format) {
+  if (isWavFormat(format) || ascii(new Uint8Array(buffer), 0, "RIFF")) {
+    const parsed = parseWav(buffer);
+    if (parsed.format.byteLength < 16) throw new Error("Invalid WAV fmt chunk.");
+    const view = new DataView(parsed.format.buffer, parsed.format.byteOffset, parsed.format.byteLength);
+    const sampleRate = view.getUint32(4, true);
+    const channels = view.getUint16(2, true);
+    const bitsPerSample = parsed.format.byteLength >= 16 ? view.getUint16(14, true) : 0;
+    const formatCode = view.getUint16(0, true);
+    return {
+      format,
+      mimeType: "audio/wav",
+      codec: formatCode === 1 ? "pcm" : "unknown",
+      sampleRate,
+      channels,
+      ...sampleRate && channels && bitsPerSample ? { bitrate: sampleRate * channels * bitsPerSample } : {},
+      isCompressed: formatCode !== 1
+    };
+  }
+  if (isMp3Format(format)) return parseMp3Specification(buffer, format) ?? formatAudioSpecification(format);
+  return formatAudioSpecification(format);
+}
+function validateAudioSpecifications(specs) {
+  const first = specs[0];
+  if (!first) return;
+  const mismatch = specs.find(
+    (spec) => spec.sampleRate !== first.sampleRate || spec.channels !== first.channels || first.bitrate !== void 0 && spec.bitrate !== void 0 && spec.bitrate !== first.bitrate
+  );
+  if (mismatch)
+    throw new AudioFormatMismatchError(
+      `Audio chunks have incompatible specifications: ${first.sampleRate}Hz/${first.channels}ch versus ${mismatch.sampleRate}Hz/${mismatch.channels}ch.`,
+      specs
+    );
+}
+function isAudioFormatMismatch(error) {
+  return error instanceof AudioFormatMismatchError || error !== null && typeof error === "object" && "kind" in error && error.kind === "audio-format-mismatch";
+}
+function writeUint32(target, offset, value) {
+  new DataView(target.buffer).setUint32(offset, value, true);
+}
+function writeChunk(target, offset, id, data) {
+  for (let index = 0; index < 4; index += 1) target[offset + index] = id.charCodeAt(index) ?? 0;
+  writeUint32(target, offset + 4, data.byteLength);
+  target.set(data, offset + 8);
+  const end = offset + 8 + data.byteLength;
+  if (data.byteLength & 1) target[end] = 0;
+  return end + (data.byteLength & 1);
+}
+function mergeWavBuffers(buffers) {
+  if (buffers.length === 0) return new ArrayBuffer(0);
+  const parsed = buffers.map(parseWav);
+  const first = parsed[0];
+  if (!first) throw new Error("At least one WAV buffer is required.");
+  if (parsed.some(
+    (item) => item.format.length !== first.format.length || item.format.some((value, i) => value !== first.format[i])
+  ))
+    throw new Error("WAV buffers have incompatible fmt chunks.");
+  const dataLength = parsed.reduce((total, item) => total + item.data.byteLength, 0);
+  const nonDataLength = first.chunks.reduce(
+    (total, chunk) => chunk.id === "data" ? total : total + 8 + chunk.data.byteLength + (chunk.data.byteLength & 1),
+    0
+  );
+  const outputLength = 12 + nonDataLength + 8 + dataLength + (dataLength & 1);
+  if (outputLength - 8 > 4294967295) throw new RangeError("Merged WAV exceeds the RIFF format size limit.");
+  const output = new Uint8Array(outputLength);
+  output.set(Uint8Array.from([82, 73, 70, 70]), 0);
+  writeUint32(output, 4, outputLength - 8);
+  output.set(Uint8Array.from([87, 65, 86, 69]), 8);
+  let outputOffset = 12;
+  let dataWritten = false;
+  for (const chunk of first.chunks) {
+    if (chunk.id === "data") {
+      if (dataWritten) continue;
+      const data = new Uint8Array(dataLength);
+      let dataOffset = 0;
+      for (const item of parsed) {
+        data.set(item.data, dataOffset);
+        dataOffset += item.data.byteLength;
+      }
+      outputOffset = writeChunk(output, outputOffset, "data", data);
+      dataWritten = true;
+    } else {
+      outputOffset = writeChunk(output, outputOffset, chunk.id, chunk.data);
     }
-    if (token.parentName?.toLowerCase() !== "speak" || token.childElementIndex !== 0)
-      addDiagnostic2(
-        diagnostics,
-        source,
-        token.start,
-        "<mstts:backgroundaudio> must be the first element directly under <speak>."
-      );
-    if (!token.selfClosing)
-      addDiagnostic2(diagnostics, source, token.start, "<mstts:backgroundaudio> must be self-closing.");
   }
+  if (!dataWritten) throw new Error("WAV audio must contain a data chunk.");
+  return output.buffer;
 }
-function validateAzureSsmlStatic2(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"
+function skipId3v2(bytes) {
+  if (!ascii(bytes, 0, "ID3") || bytes.byteLength < 10) return 0;
+  const size = [bytes[6], bytes[7], bytes[8], bytes[9]].reduce((total, value) => total << 7 | value & 127, 0);
+  const hasFooter = (bytes[5] & 16) !== 0;
+  return Math.min(bytes.byteLength, 10 + size + (hasFooter ? 10 : 0));
+}
+function stripMp3Tags(buffer) {
+  const bytes = new Uint8Array(buffer);
+  const start = skipId3v2(bytes);
+  const end = bytes.byteLength >= 128 && ascii(bytes, bytes.byteLength - 128, "TAG") ? bytes.byteLength - 128 : bytes.byteLength;
+  return bytes.slice(Math.min(start, end), end);
+}
+function isMp3Format(format) {
+  return /(?:mp3|mpeg)/i.test(format);
+}
+function isWavFormat(format) {
+  return /(?:wav|wave|riff)/i.test(format);
+}
+function isRawFormat(format) {
+  return /^raw(?:-|$)/i.test(format);
+}
+function resolveMergeAudioFormat(format) {
+  if (isWavFormat(format)) return "wav";
+  if (isMp3Format(format)) return "mp3";
+  if (isRawFormat(format)) return "raw";
+  return void 0;
+}
+function canMergeAudioFormat(format) {
+  return resolveMergeAudioFormat(format) !== void 0;
+}
+function mergeAudioBuffers(buffers, options) {
+  const format = typeof options === "string" ? options : options?.format;
+  if (!format) throw new UnsupportedMergeFormatError("");
+  try {
+    validateAudioSpecifications(buffers.map((buffer) => inspectAudioSpecification(buffer, format)));
+    if (isWavFormat(format)) return mergeWavBuffers(buffers);
+    if (isMp3Format(format)) {
+      const parts = buffers.map(stripMp3Tags);
+      const output = new Uint8Array(parts.reduce((total, part) => total + part.byteLength, 0));
+      let offset = 0;
+      for (const part of parts) {
+        output.set(part, offset);
+        offset += part.byteLength;
       }
-    ];
+      return output.buffer;
+    }
+    if (isRawFormat(format)) {
+      const output = new Uint8Array(buffers.reduce((total, buffer) => total + buffer.byteLength, 0));
+      let offset = 0;
+      for (const buffer of buffers) {
+        output.set(new Uint8Array(buffer), offset);
+        offset += buffer.byteLength;
+      }
+      return output.buffer;
+    }
+    throw new UnsupportedMergeFormatError(format);
+  } catch (error) {
+    if (error instanceof UnsupportedMergeFormatError || isAudioFormatMismatch(error) || error instanceof MergeError)
+      throw error;
+    throw new MergeError(`Audio buffers could not be merged for format "${format}".`, error);
   }
-  const maxLength = options.maxLength ?? 1e4;
-  if (ssml.length > maxLength)
-    addDiagnostic2(diagnostics, ssml, maxLength, `SSML exceeds the maximum length of ${maxLength} characters.`);
-  if (options.maxXmlDepth !== void 0 && (!Number.isInteger(options.maxXmlDepth) || options.maxXmlDepth <= 0)) {
-    addDiagnostic2(diagnostics, ssml, 0, "maxXmlDepth must be a positive integer.");
+}
+function closeSpeechResources(speechConfig, synthesizer) {
+  try {
+    synthesizer.close();
+  } catch {
   }
   try {
-    parseSsml2(ssml);
-  } catch (error) {
-    const message = error instanceof Error ? error.message.replace(/ at position \d+$/, "") : String(error);
-    const match = / at position (\d+)$/.exec(error instanceof Error ? error.message : "");
-    addDiagnostic2(diagnostics, ssml, match ? Number(match[1]) : 0, message);
-    return diagnostics;
+    speechConfig.close();
+  } catch {
   }
-  const tokens = tokenizeElements2(ssml);
-  if (options.maxXmlDepth !== void 0) {
-    for (const token of tokens) {
-      if (token.depth > options.maxXmlDepth) {
-        addDiagnostic2(
-          diagnostics,
-          ssml,
-          token.start,
-          `XML nesting depth ${token.depth} exceeds the configured maximum of ${options.maxXmlDepth}.`
-        );
+}
+var ticksToMilliseconds = (ticks) => Math.max(0, ticks) / 1e4;
+async function synthesizeSsml(ssml, config) {
+  if (config.signal?.aborted) {
+    throw new SynthesisCancelledError();
+  }
+  const speechConfig = createSpeechConfig(config);
+  const synthesizer = new SpeechSDK2.SpeechSynthesizer(speechConfig, null);
+  return await new Promise((resolve, reject) => {
+    let resourcesClosed = false;
+    let settled = false;
+    let timeout;
+    let abortHandler;
+    const cleanup = () => {
+      if (timeout) clearTimeout(timeout);
+      if (abortHandler) config.signal?.removeEventListener("abort", abortHandler);
+    };
+    const closeResources = () => {
+      if (resourcesClosed) return;
+      resourcesClosed = true;
+      closeSpeechResources(speechConfig, synthesizer);
+    };
+    const rejectWithError = (error) => {
+      if (settled) return;
+      settled = true;
+      cleanup();
+      closeResources();
+      reject(toSynthesisError(error));
+    };
+    const boundaries = [];
+    const visemes = [];
+    const bookmarks = [];
+    let sourceEventCursor = 0;
+    let generatedSourceMap;
+    if (!config.sourceTextSegments && !config.sourceMarkers) {
+      try {
+        generatedSourceMap = getSsmlSourceMap2(ssml);
+      } catch {
+        generatedSourceMap = void 0;
       }
     }
-  }
-  const speak = tokens.find((token) => token.name.toLowerCase() === "speak");
-  const voices = tokens.filter((token) => token.name.toLowerCase() === "voice");
-  const backgroundAudioTokens = tokens.filter((token) => token.name.toLowerCase() === "mstts:backgroundaudio");
-  for (const [index, token] of backgroundAudioTokens.entries()) {
-    if (index > 0)
-      addDiagnostic2(
-        diagnostics,
-        ssml,
-        token.start,
-        "An SSML document can contain at most one <mstts:backgroundaudio> element."
-      );
-  }
-  if (!speak || voices.length === 0)
-    addDiagnostic2(
-      diagnostics,
-      ssml,
-      speak?.start ?? 0,
-      "Azure SSML requires at least one <voice> element under <speak>."
-    );
-  const voiceName = voices[0] ? attr2(voices[0], "name") : void 0;
-  const voiceCatalog = normalizeVoiceCatalog2(options);
-  const normalizeLanguage = createLanguageNormalizer2(options);
-  const policySeverity = diagnosticSeverity2(options.unknownVoicePolicy ?? "warn");
-  const voicesToValidate = options.validateNestedVoices === false ? voices.slice(0, 1) : voices;
-  for (const token of voicesToValidate) {
-    const name = attr2(token, "name")?.trim();
-    const language = attr2(token, "xml:lang")?.trim() || (speak ? attr2(speak, "xml:lang")?.trim() : void 0);
-    const definition = name ? voiceCatalog.get(name.toLowerCase()) : void 0;
-    if (name && definition?.status === "preview")
-      addDiagnostic2(
-        diagnostics,
-        ssml,
-        token.start,
-        `Voice "${name}" is an Azure Speech preview voice and may change or require preview access.`,
-        "warning",
-        "azure-preview-voice"
-      );
-    if (name && definition?.status === "deprecated")
-      addDiagnostic2(
-        diagnostics,
-        ssml,
-        token.start,
-        `Voice "${name}" is deprecated by Azure Speech; migrate to a supported voice.`,
-        "info",
-        "azure-deprecated-voice"
-      );
-    if (name && !definition && policySeverity)
-      addDiagnostic2(
-        diagnostics,
-        ssml,
-        token.start,
-        `Unknown voice "${name}" is not registered in the voice catalog.`,
-        policySeverity,
-        "azure-unknown-voice"
-      );
-    if (name && language && definitionMatchesLanguage2(definition, name, language, normalizeLanguage) === false)
-      addDiagnostic2(
-        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" ? attr2(token, "name")?.trim() : tokenName === "mstts:turn" ? attr2(token, "voice")?.trim() || token.parentVoiceName : options.validateNestedVoices === false ? voiceName : token.parentVoiceName;
-    validateElement2(token, ssml, diagnostics, tokenVoiceName, options, voiceCatalog);
-    const definition = tokenVoiceName ? voiceCatalog.get(tokenVoiceName.toLowerCase()) : void 0;
-    validateVoiceFeatureMatrix2(token, ssml, diagnostics, tokenVoiceName, definition);
-    if (tokenName === "voice" && options.model && definition?.models && !definition.models.some((model) => model.toLowerCase() === options.model?.toLowerCase())) {
-      addDiagnostic2(
-        diagnostics,
-        ssml,
-        token.start,
-        `Voice "${tokenVoiceName}" does not support model "${options.model}" according to the configured feature matrix.`,
-        "error",
-        "azure-unsupported-model-for-voice"
+    const sourceBaseOffset = config.sourceTextRange?.start ?? 0;
+    const sourceSegments = config.sourceTextSegments ?? generatedSourceMap?.segments.map((segment) => ({
+      ...segment,
+      range: {
+        start: segment.range.start + sourceBaseOffset,
+        end: segment.range.end + sourceBaseOffset
+      },
+      sourceNodePath: [...segment.sourceNodePath]
+    })) ?? [];
+    const sourceMarkers = config.sourceMarkers ?? generatedSourceMap?.markers.map((marker) => ({
+      ...marker,
+      originalTextRange: {
+        start: marker.originalTextRange.start + sourceBaseOffset,
+        end: marker.originalTextRange.end + sourceBaseOffset
+      },
+      sourceNodePath: [...marker.sourceNodePath]
+    })) ?? [];
+    const sourceText = sourceSegments.map((segment) => segment.text).join("");
+    const mapSourceEvent = (text, offsetHint, markerName) => {
+      const marker = markerName ? sourceMarkers.find((candidate) => candidate.name === markerName) : void 0;
+      if (marker) {
+        return {
+          originalTextRange: { ...marker.originalTextRange },
+          sourceNodePath: [...marker.sourceNodePath],
+          textRange: { ...marker.originalTextRange },
+          mappingStatus: "exact"
+        };
+      }
+      if (sourceSegments.length === 0 && !config.sourceTextRange && !config.sourceNodePath) {
+        const unmapped = { mappingStatus: "unmapped" };
+        Object.defineProperty(unmapped, "mappingStatus", { value: "unmapped", enumerable: false });
+        return unmapped;
+      }
+      const value = text ?? "";
+      let localStart = Number.isFinite(offsetHint) && (offsetHint ?? 0) >= 0 ? offsetHint : -1;
+      let mappingStatus = "exact";
+      if (value && localStart >= 0 && sourceText.slice(localStart, localStart + value.length) !== value) {
+        localStart = -1;
+        mappingStatus = "fallback";
+      }
+      if (localStart < 0 || localStart > sourceText.length) {
+        localStart = value ? sourceText.indexOf(value, sourceEventCursor) : sourceEventCursor;
+        if (localStart < 0) localStart = value ? sourceText.indexOf(value) : sourceEventCursor;
+        mappingStatus = "fallback";
+      }
+      localStart = Math.max(0, localStart);
+      const localEnd = Math.min(sourceText.length, localStart + value.length);
+      sourceEventCursor = Math.max(sourceEventCursor, localEnd);
+      const baseStart = config.sourceTextRange?.start ?? sourceSegments[0]?.range.start ?? 0;
+      const fallbackRange = { start: baseStart + localStart, end: baseStart + localEnd };
+      const segment = sourceSegments.find(({ range }) => range.start <= fallbackRange.start && range.end > fallbackRange.start) ?? sourceSegments.find(({ range }) => range.end > fallbackRange.start) ?? (value.length === 0 ? sourceSegments.find(({ range }) => range.start <= fallbackRange.start && range.end >= fallbackRange.start) : void 0);
+      return {
+        originalTextRange: { ...fallbackRange },
+        textRange: { ...fallbackRange },
+        ...segment ? { sourceNodePath: [...segment.sourceNodePath] } : config.sourceNodePath ? { sourceNodePath: [...config.sourceNodePath] } : {},
+        mappingStatus: segment || config.sourceTextRange || config.sourceNodePath ? mappingStatus : "unmapped"
+      };
+    };
+    synthesizer.wordBoundary = (_sender, event) => {
+      boundaries.push({
+        text: event.text,
+        audioOffsetMs: ticksToMilliseconds(event.audioOffset),
+        durationMs: ticksToMilliseconds(event.duration),
+        ...mapSourceEvent(
+          event.text,
+          event.textOffset
+        )
+      });
+    };
+    synthesizer.visemeReceived = (_sender, event) => {
+      const eventWithOffset = event;
+      visemes.push({
+        visemeId: event.visemeId,
+        audioOffsetMs: ticksToMilliseconds(event.audioOffset),
+        ...mapSourceEvent(void 0, eventWithOffset.textOffset)
+      });
+    };
+    synthesizer.bookmarkReached = (_sender, event) => {
+      const eventWithOffset = event;
+      bookmarks.push({
+        name: event.text,
+        audioOffsetMs: ticksToMilliseconds(event.audioOffset),
+        ...mapSourceEvent(void 0, eventWithOffset.textOffset, event.text)
+      });
+    };
+    const cb = (result) => {
+      if (settled) return;
+      const { reason, errorDetails } = result;
+      if (reason !== SpeechSDK2.ResultReason.SynthesizingAudioCompleted) {
+        const err = errorDetails || `Speech synthesis failed with reason ${reason}.`;
+        rejectWithError(err);
+        return;
+      }
+      settled = true;
+      cleanup();
+      closeResources();
+      const eventDurationMs = Math.max(
+        0,
+        ...(boundaries ?? []).map((boundary) => boundary.audioOffsetMs + boundary.durationMs),
+        ...(visemes ?? []).map((viseme) => viseme.audioOffsetMs),
+        ...(bookmarks ?? []).map((bookmark) => bookmark.audioOffsetMs)
       );
+      const durationMs = result.audioDuration ? ticksToMilliseconds(result.audioDuration) : eventDurationMs;
+      const requestId = result.resultId;
+      const addSourceMetadata = (event) => {
+        const mapped = {
+          ...event,
+          ...config.sourceTextRange && !("textRange" in event) ? { textRange: { ...config.sourceTextRange } } : {},
+          ...config.sourceTextRange && !("originalTextRange" in event) ? { originalTextRange: { ...config.sourceTextRange } } : {},
+          ...config.chunkIndex !== void 0 ? { chunkIndex: config.chunkIndex } : {},
+          ...config.sourceNodePath ? { sourceNodePath: [...config.sourceNodePath] } : {},
+          ...requestId ? { requestId } : {}
+        };
+        if (event.mappingStatus === "unmapped")
+          Object.defineProperty(mapped, "mappingStatus", { value: "unmapped", enumerable: false });
+        return mapped;
+      };
+      const sourceBoundaries = boundaries.map((boundary) => addSourceMetadata(boundary));
+      const sourceVisemes = visemes.map((viseme) => addSourceMetadata(viseme));
+      const sourceBookmarks = bookmarks.map((bookmark) => addSourceMetadata(bookmark));
+      resolve({
+        audioData: result.audioData,
+        durationMs,
+        audioSpec: inspectAudioSpecification(result.audioData, config.outputFormat ?? DEFAULT_OUTPUT_FORMAT),
+        mimeType: resolveMimeType(config.outputFormat ?? DEFAULT_OUTPUT_FORMAT),
+        ...config.sourceTextRange ? { textRange: { ...config.sourceTextRange } } : {},
+        ...requestId ? { requestId } : {},
+        ...sourceBoundaries.length > 0 ? { boundaries: sourceBoundaries, wordBoundary: sourceBoundaries, wordBoundaries: sourceBoundaries } : {},
+        ...sourceVisemes.length > 0 ? { visemes: sourceVisemes } : {},
+        ...sourceBookmarks.length > 0 ? { bookmarks: sourceBookmarks } : {}
+      });
+    };
+    try {
+      if (config.signal) {
+        abortHandler = () => rejectWithError(new SynthesisCancelledError());
+        config.signal.addEventListener("abort", abortHandler, { once: true });
+      }
+      if (config.timeoutMs !== void 0 && config.timeoutMs > 0) {
+        timeout = setTimeout(
+          () => rejectWithError(new SynthesisTimeoutError(`Speech synthesis timed out after ${config.timeoutMs} ms.`)),
+          config.timeoutMs
+        );
+      }
+      synthesizer.speakSsmlAsync(ssml, cb, rejectWithError);
+    } catch (error) {
+      rejectWithError(error);
     }
-  }
-  return diagnostics;
-}
-function urlAttributes2(token) {
-  const tag = canonicalTagName2(token.name);
-  const attributes = tag === "audio" || tag === "mstts:backgroundaudio" ? ["src"] : tag === "lexicon" ? ["uri"] : tag === "mstts:voiceconversion" ? ["url"] : [];
-  return attributes.flatMap((attribute) => {
-    const value = attr2(token, attribute);
-    return value === void 0 ? [] : [{ attribute, value }];
   });
 }
-function validateAzureSsml2(ssml, options = {}) {
-  const diagnostics = validateAzureSsmlStatic2(ssml, options);
-  const validator = options.urlValidator ?? options.customUrlValidator;
-  if (!validator || typeof ssml !== "string") return diagnostics;
-  const runnerOptions = options.urlValidation ?? {};
-  const boundedValidator = createAzureUrlValidatorRunner2(validator, {
-    ...runnerOptions,
-    ...options.urlValidatorConcurrency !== void 0 ? { concurrency: options.urlValidatorConcurrency } : {},
-    ...options.urlValidatorTimeoutMs !== void 0 ? { timeoutMs: options.urlValidatorTimeoutMs } : {},
-    ...options.urlValidatorSignal ? { signal: options.urlValidatorSignal } : {},
-    ...options.urlValidatorCache ? { cache: options.urlValidatorCache } : {}
+function isRetryableSynthesisError(error) {
+  if (error instanceof SynthesisCancelledError || error instanceof SynthesisTimeoutError) return false;
+  if (error instanceof AzureTtsError && error.status !== 0)
+    return error.status === 429 || error.status >= 500 && error.status < 600;
+  const message = error instanceof Error ? error.message : String(error);
+  if (/\b4\d{2}\b/.test(message)) return false;
+  const status = error && typeof error === "object" && "status" in error ? error.status : void 0;
+  if (typeof status === "number") return status === 429 || status >= 500 && status < 600;
+  return /network|connection|connect|socket|fetch failed|econn|etimedout|temporar|transient|unavailable/i.test(message);
+}
+function retryDelay(options, retryAttempt) {
+  const base = Math.min(options.maxDelayMs, options.initialDelayMs * 2 ** Math.max(0, retryAttempt - 1));
+  return Math.floor(Math.random() * (base + 1));
+}
+function resolveConcurrency(value, total) {
+  if (value === void 0) return 1;
+  if (value === Infinity) return Math.max(1, total);
+  return Number.isFinite(value) ? Math.max(1, Math.floor(value)) : 1;
+}
+async function waitForRetry(delayMs, signal) {
+  if (signal?.aborted) throw new SynthesisCancelledError();
+  if (delayMs <= 0) return;
+  await new Promise((resolve, reject) => {
+    let timer;
+    const abort = () => {
+      clearTimeout(timer);
+      signal?.removeEventListener("abort", abort);
+      reject(new SynthesisCancelledError());
+    };
+    timer = setTimeout(() => {
+      signal?.removeEventListener("abort", abort);
+      resolve();
+    }, delayMs);
+    if (signal) {
+      signal.addEventListener("abort", abort, { once: true });
+    }
   });
-  let tokens;
-  try {
-    tokens = tokenizeElements2(ssml);
-  } catch {
-    return diagnostics;
+}
+async function synthesizeWithRetry(ssml, config, retryOptions, onRetry) {
+  const options = retryOptions ? {
+    maxRetries: Math.max(0, Math.floor(retryOptions.maxRetries)),
+    initialDelayMs: Math.max(0, retryOptions.initialDelayMs),
+    maxDelayMs: Math.max(0, retryOptions.maxDelayMs)
+  } : void 0;
+  let attempt = 0;
+  while (true) {
+    if (config.signal?.aborted) throw new SynthesisCancelledError();
+    try {
+      return await synthesizeSsml(ssml, config);
+    } catch (error) {
+      if (!options || attempt >= options.maxRetries || !isRetryableSynthesisError(error)) throw error;
+      attempt += 1;
+      const delayMs = retryDelay(options, attempt);
+      onRetry(attempt, delayMs);
+      await waitForRetry(delayMs, config.signal);
+    }
   }
-  const checks = tokens.flatMap(
-    (token) => urlAttributes2(token).map(async ({ attribute, value }) => {
+}
+async function synthesizeSsmlChunks(chunks, config) {
+  const results = new Array(chunks.length);
+  const totalChunks = chunks.length;
+  const report = (event) => config.onProgress?.(event);
+  for (const [index, chunk] of chunks.entries()) {
+    const input = typeof chunk === "string" ? { ssml: chunk } : chunk;
+    report({
+      currentChunk: index,
+      totalChunks,
+      percent: totalChunks === 0 ? 100 : Math.round(index / totalChunks * 100),
+      chunkIndex: index,
+      originalTextRange: input.originalTextRange,
+      status: "pending",
+      durationMs: 0
+    });
+  }
+  let completed = 0;
+  let nextIndex = 0;
+  const concurrency = resolveConcurrency(config.concurrency, chunks.length);
+  const worker = async () => {
+    while (true) {
+      const index = nextIndex++;
+      if (index >= chunks.length) return;
+      const chunk = chunks[index];
+      const input = typeof chunk === "string" ? { ssml: chunk } : chunk;
+      report({
+        currentChunk: completed,
+        totalChunks,
+        percent: totalChunks === 0 ? 100 : Math.round(completed / totalChunks * 100),
+        chunkIndex: index,
+        originalTextRange: input.originalTextRange,
+        status: "synthesizing",
+        durationMs: 0
+      });
+      const startedAt = Date.now();
       try {
-        const result = await boundedValidator(value, { tag: token.name, attribute });
-        const valid = typeof result === "boolean" ? result : result.valid;
-        if (!valid) {
-          const reason = typeof result === "boolean" ? void 0 : result.reason;
-          addDiagnostic2(
-            diagnostics,
-            ssml,
-            token.start,
-            `<${token.name} ${attribute}> was rejected by the custom URL validator${reason ? `: ${reason}` : "."}`
-          );
-        }
-      } catch (error) {
-        const reason = error instanceof Error ? error.message : String(error);
-        addDiagnostic2(
-          diagnostics,
-          ssml,
-          token.start,
-          `<${token.name} ${attribute}> could not be validated by the custom URL validator: ${reason}`
+        const result = await synthesizeWithRetry(
+          input.ssml,
+          {
+            ...config,
+            ...input.originalTextRange ? { sourceTextRange: input.originalTextRange } : {},
+            ...input.sourceNodePath ?? config.sourceNodePath ? { sourceNodePath: [...input.sourceNodePath ?? config.sourceNodePath ?? []] } : {},
+            ...input.sourceTextSegments ? { sourceTextSegments: input.sourceTextSegments } : {},
+            ...input.sourceMarkers ? { sourceMarkers: input.sourceMarkers } : {},
+            chunkIndex: index,
+            onProgress: void 0
+          },
+          config.retryOptions,
+          (retryAttempt, nextRetryDelayMs) => report({
+            currentChunk: completed,
+            totalChunks,
+            percent: totalChunks === 0 ? 100 : Math.round(completed / totalChunks * 100),
+            chunkIndex: index,
+            originalTextRange: input.originalTextRange,
+            status: "synthesizing",
+            durationMs: Date.now() - startedAt,
+            retryAttempt,
+            nextRetryDelayMs,
+            isRetrying: true
+          })
         );
+        results[index] = result;
+        completed += 1;
+        report({
+          currentChunk: completed,
+          totalChunks,
+          percent: totalChunks === 0 ? 100 : Math.round(completed / totalChunks * 100),
+          chunkIndex: index,
+          originalTextRange: input.originalTextRange,
+          status: "success",
+          durationMs: Date.now() - startedAt
+        });
+      } catch (error) {
+        report({
+          currentChunk: completed,
+          totalChunks,
+          percent: totalChunks === 0 ? 100 : Math.round(completed / totalChunks * 100),
+          chunkIndex: index,
+          originalTextRange: input.originalTextRange,
+          status: "failed",
+          durationMs: Date.now() - startedAt,
+          error
+        });
+        throw error;
       }
-    })
-  );
-  return Promise.all(checks).then(() => diagnostics);
+    }
+  };
+  await Promise.all(Array.from({ length: Math.min(concurrency, chunks.length) }, () => worker()));
+  const orderedResults = results.filter((result) => result !== void 0);
+  return mergeSynthesisResults(orderedResults, {
+    format: config.outputFormat ?? DEFAULT_OUTPUT_FORMAT,
+    signal: config.signal
+  });
+}
+function createMergedResult(results, audioData, format, audioSpec, outputMimeType) {
+  const boundaries = [];
+  const visemes = [];
+  const bookmarks = [];
+  let durationOffset = 0;
+  for (const [resultIndex, result] of results.entries()) {
+    const chunkBoundaries = result.boundaries && result.boundaries.length > 0 ? result.boundaries : result.wordBoundary ?? result.wordBoundaries ?? [];
+    for (const boundary of chunkBoundaries) {
+      const textRange = boundary.textRange ?? result.textRange;
+      const originalTextRange = boundary.originalTextRange ?? textRange;
+      const requestId = boundary.requestId ?? result.requestId;
+      boundaries.push({
+        ...boundary,
+        audioOffsetMs: boundary.audioOffsetMs + durationOffset,
+        chunkAudioOffsetMs: boundary.chunkAudioOffsetMs ?? boundary.audioOffsetMs,
+        ...boundary.chunkIndex === void 0 ? { chunkIndex: resultIndex } : {},
+        ...boundary.sourceNodePath ? { sourceNodePath: [...boundary.sourceNodePath] } : {},
+        ...originalTextRange ? { originalTextRange: { ...originalTextRange } } : {},
+        ...textRange ? { textRange: { ...textRange } } : {},
+        ...requestId ? { requestId } : {},
+        mappingStatus: boundary.mappingStatus ?? "unmapped"
+      });
+    }
+    for (const viseme of result.visemes ?? []) {
+      const textRange = viseme.textRange ?? result.textRange;
+      const originalTextRange = viseme.originalTextRange ?? textRange;
+      const requestId = viseme.requestId ?? result.requestId;
+      visemes.push({
+        ...viseme,
+        audioOffsetMs: viseme.audioOffsetMs + durationOffset,
+        chunkAudioOffsetMs: viseme.chunkAudioOffsetMs ?? viseme.audioOffsetMs,
+        ...viseme.chunkIndex === void 0 ? { chunkIndex: resultIndex } : {},
+        ...viseme.sourceNodePath ? { sourceNodePath: [...viseme.sourceNodePath] } : {},
+        ...originalTextRange ? { originalTextRange: { ...originalTextRange } } : {},
+        ...textRange ? { textRange: { ...textRange } } : {},
+        ...requestId ? { requestId } : {},
+        mappingStatus: viseme.mappingStatus ?? "unmapped"
+      });
+    }
+    for (const bookmark of result.bookmarks ?? []) {
+      const textRange = bookmark.textRange ?? result.textRange;
+      const originalTextRange = bookmark.originalTextRange ?? textRange;
+      const requestId = bookmark.requestId ?? result.requestId;
+      bookmarks.push({
+        ...bookmark,
+        audioOffsetMs: bookmark.audioOffsetMs + durationOffset,
+        chunkAudioOffsetMs: bookmark.chunkAudioOffsetMs ?? bookmark.audioOffsetMs,
+        ...bookmark.chunkIndex === void 0 ? { chunkIndex: resultIndex } : {},
+        ...bookmark.sourceNodePath ? { sourceNodePath: [...bookmark.sourceNodePath] } : {},
+        ...originalTextRange ? { originalTextRange: { ...originalTextRange } } : {},
+        ...textRange ? { textRange: { ...textRange } } : {},
+        ...requestId ? { requestId } : {},
+        mappingStatus: bookmark.mappingStatus ?? "unmapped"
+      });
+    }
+    durationOffset += Math.max(0, result.durationMs);
+  }
+  return {
+    audioData,
+    durationMs: durationOffset,
+    mimeType: resolveMimeType(format),
+    audioSpec: audioSpec ?? formatAudioSpecification(format),
+    ...outputMimeType ? { mimeType: outputMimeType } : {},
+    ...boundaries.length > 0 ? { boundaries, wordBoundary: boundaries, wordBoundaries: boundaries } : {},
+    ...visemes.length > 0 ? { visemes } : {},
+    ...bookmarks.length > 0 ? { bookmarks } : {},
+    ...results.length === 1 && results[0]?.requestId ? { requestId: results[0].requestId } : {},
+    ...results.length === 1 && results[0]?.textRange ? { textRange: { ...results[0].textRange } } : {}
+  };
+}
+function mergeSynthesisResults(results, options) {
+  const resolvedOptions = typeof options === "string" ? { format: options } : options;
+  const format = resolvedOptions?.format;
+  if (!format) throw new UnsupportedMergeFormatError("");
+  const buffers = results.map((result) => result.audioData);
+  const inputSpecs = results.map((result) => result.audioSpec ?? inspectAudioSpecification(result.audioData, format));
+  validateAudioSpecifications(inputSpecs);
+  const signal = resolvedOptions.signal ?? new AbortController().signal;
+  if (signal.aborted) throw new SynthesisCancelledError();
+  if (resolvedOptions.customMerger) {
+    return Promise.resolve().then(
+      () => resolvedOptions.customMerger?.(buffers, {
+        format,
+        outputMimeType: resolvedOptions.outputMimeType ?? resolveMimeType(format),
+        inputSpecs,
+        signal
+      })
+    ).then((merged) => {
+      if (!merged) throw new MergeError("The custom audio merger returned no audio buffer.");
+      if (!(merged instanceof ArrayBuffer) || buffers.some((buffer) => buffer.byteLength > 0) && merged.byteLength === 0)
+        throw new MergeError("The custom audio merger returned an invalid audio buffer.");
+      if (signal.aborted) throw new SynthesisCancelledError();
+      return createMergedResult(
+        results,
+        merged,
+        format,
+        inspectAudioSpecification(merged, format),
+        resolvedOptions.outputMimeType
+      );
+    }).catch((error) => {
+      if (error instanceof UnsupportedMergeFormatError || isAudioFormatMismatch(error) || error instanceof MergeError)
+        throw error;
+      throw new MergeError(`Custom audio merger failed for format "${format}".`, error);
+    });
+  }
+  try {
+    return createMergedResult(
+      results,
+      mergeAudioBuffers(buffers, { format }),
+      format,
+      inputSpecs[0],
+      resolvedOptions.outputMimeType
+    );
+  } catch (error) {
+    if (error instanceof UnsupportedMergeFormatError || isAudioFormatMismatch(error) || error instanceof MergeError)
+      throw error;
+    throw new MergeError(`Audio buffers could not be merged for format "${format}".`, error);
+  }
+}
+async function synthesizeSpeech(ssml, config) {
+  return (await synthesizeSsml(ssml, config)).audioData;
 }
-var AZURE_VOICE_CATALOG_METADATA2 = {
-  apiVersion: "2025-10-01",
-  generatedAt: "2026-08-28T00:00:00.000Z",
-  regions: [],
-  voiceCount: AZURE_VOICE_DEFINITIONS2.length
-};
 
 // packages/azure-tts-client/src/safe.ts
 var ChunkValidationError = class extends Error {
   constructor(chunkIndex, diagnostics) {
     super(`SSML validation failed for chunk ${chunkIndex}; the Azure Speech API was not called.`);
-    this.kind = "chunk-validation";
+    this.kind = "validation-error";
     this.name = "ChunkValidationError";
     this.chunkIndex = chunkIndex;
     this.diagnostics = diagnostics;
   }
 };
+function failure(error) {
+  return { ok: false, success: false, status: error.kind, error };
+}
+function isRetryable(error) {
+  if (error instanceof Error && /cancel|abort|tim(?:e|ed) ?out/i.test(error.message)) return false;
+  const status = error && typeof error === "object" && "status" in error ? error.status : void 0;
+  if (typeof status === "number" && status !== 0) return status === 429 || status >= 500 && status < 600;
+  const message = error instanceof Error ? error.message : String(error);
+  if (/\b4\d{2}\b/.test(message)) return false;
+  return /network|connection|connect|socket|fetch failed|econn|etimedout|temporar|transient|unavailable/i.test(message);
+}
+function delayForRetry(options, attempt) {
+  const maxDelay = Math.max(0, options.maxDelayMs);
+  const base = Math.min(maxDelay, Math.max(0, options.initialDelayMs) * 2 ** Math.max(0, attempt - 1));
+  return Math.floor(Math.random() * (base + 1));
+}
+function resolveConcurrency2(value, total) {
+  if (value === void 0) return 1;
+  if (value === Infinity) return Math.max(1, total);
+  return Number.isFinite(value) ? Math.max(1, Math.floor(value)) : 1;
+}
+async function retryableSynthesis(synthesize, options, signal, onRetry) {
+  const retry = options ? {
+    maxRetries: Math.max(0, Math.floor(options.maxRetries)),
+    initialDelayMs: options.initialDelayMs,
+    maxDelayMs: options.maxDelayMs
+  } : void 0;
+  let attempt = 0;
+  while (true) {
+    if (signal?.aborted) throw new Error("Speech synthesis was cancelled.");
+    try {
+      return await synthesize();
+    } catch (error) {
+      if (!retry || attempt >= retry.maxRetries || !isRetryable(error)) throw error;
+      attempt += 1;
+      const delayMs = delayForRetry(retry, attempt);
+      onRetry(attempt, delayMs);
+      if (delayMs > 0)
+        await new Promise((resolve, reject) => {
+          const timer = setTimeout(() => {
+            signal?.removeEventListener("abort", abort);
+            resolve();
+          }, delayMs);
+          const abort = () => {
+            clearTimeout(timer);
+            signal?.removeEventListener("abort", abort);
+            reject(new Error("Speech synthesis was cancelled."));
+          };
+          signal?.addEventListener("abort", abort, { once: true });
+        });
+    }
+  }
+}
+function sharedValidationOptions(options, signal) {
+  const validator = options.urlValidator ?? options.customUrlValidator;
+  if (!validator) return signal ? withValidationSignal(options, signal) : options;
+  const runner = createAzureUrlValidatorRunner2(validator, {
+    ...options.urlValidation ?? {},
+    ...options.urlValidatorConcurrency !== void 0 ? { concurrency: options.urlValidatorConcurrency } : {},
+    ...options.urlValidatorTimeoutMs !== void 0 ? { timeoutMs: options.urlValidatorTimeoutMs } : {},
+    ...signal ? { signal } : {},
+    ...options.urlValidatorCache ? { cache: options.urlValidatorCache } : {}
+  });
+  return {
+    ...withValidationSignal(options, signal),
+    urlValidatorRunner: runner
+  };
+}
 async function synthesizeSsmlSafe(client, ssml, options = {}) {
-  const validationOptions = options.validation ?? options;
+  const validationOptions = sharedValidationOptions(options.validation ?? options, options.signal);
   const diagnostics = await Promise.resolve(validateAzureSsml2(ssml, validationOptions));
+  if (options.signal?.aborted) {
+    const error = toSynthesisError(new Error("Speech synthesis was cancelled."));
+    return failure(error);
+  }
   const errors = diagnostics.filter((diagnostic) => diagnostic.severity === "error");
   if (errors.length > 0) {
-    return {
-      ok: false,
-      success: false,
-      status: "validation-error",
-      error: {
-        kind: "validation",
-        message: "SSML validation failed; the Azure Speech API was not called.",
-        diagnostics: errors
-      }
-    };
+    return failure({
+      kind: "validation-error",
+      message: "SSML validation failed; the Azure Speech API was not called.",
+      diagnostics: errors
+    });
   }
   try {
-    return { ok: true, success: true, status: "success", value: await client.synthesizeSsml(ssml) };
+    return {
+      ok: true,
+      success: true,
+      status: "success",
+      value: await client.synthesizeSsml(ssml, { signal: options.signal })
+    };
   } catch (error) {
-    const azureError = error instanceof AzureTtsError ? error : createSpeechSdkError(error);
-    return { ok: false, success: false, status: "azure-api-error", error: azureError };
+    const synthesisError = toSynthesisError(error);
+    return failure(synthesisError);
   }
 }
 async function synthesizeSsmlChunksSafe(client, chunks, options = {}) {
-  const validationOptions = options.validation ?? options;
+  const validationOptions = sharedValidationOptions(options.validation ?? options, options.signal);
+  if (options.signal?.aborted) {
+    const error = toSynthesisError(new Error("Speech synthesis was cancelled."));
+    return failure(error);
+  }
   const pending = (index, status, error) => {
     options.onProgress?.({
       currentChunk: status === "success" ? index + 1 : index,
@@ -4620,77 +5428,175 @@ async function synthesizeSsmlChunksSafe(client, chunks, options = {}) {
     pending(index, "pending");
   });
   const validations = await Promise.all(
-    chunks.map(async (chunk) => {
+    chunks.map(async (chunk, index) => {
       const ssml = typeof chunk === "string" ? chunk : chunk.ssml;
-      const diagnostics = await Promise.resolve(validateAzureSsml2(ssml, validationOptions));
+      const sourceNodePath = typeof chunk === "string" ? options.sourceNodePath : chunk.sourceNodePath ?? options.sourceNodePath;
+      const diagnostics = await Promise.resolve(
+        validateAzureSsml2(ssml, {
+          ...validationOptions,
+          ...sourceNodePath ? { sourceNodePath } : {},
+          chunkIndex: index
+        })
+      );
       return diagnostics.filter((diagnostic) => diagnostic.severity === "error");
     })
   );
   const firstInvalidIndex = validations.findIndex((diagnostics) => diagnostics.length > 0);
+  if (options.signal?.aborted) {
+    const error = toSynthesisError(new Error("Speech synthesis was cancelled."));
+    return failure(error);
+  }
   if (firstInvalidIndex >= 0) {
     const error = new ChunkValidationError(firstInvalidIndex, validations[firstInvalidIndex] ?? []);
     pending(firstInvalidIndex, "failed", error);
-    return { ok: false, success: false, status: "validation-error", error };
+    return failure(error);
   }
   try {
     if (client.synthesizeChunks) {
-      const value = await client.synthesizeChunks(chunks, { onProgress: options.onProgress });
+      const normalizedChunks = chunks.map((chunk) => {
+        if (typeof chunk === "string" || chunk.sourceNodePath || !options.sourceNodePath) return chunk;
+        return { ...chunk, sourceNodePath: [...options.sourceNodePath] };
+      });
+      const value = await client.synthesizeChunks(normalizedChunks, {
+        onProgress: options.onProgress,
+        outputFormat: options.outputFormat,
+        signal: options.signal,
+        timeoutMs: options.timeoutMs,
+        sourceNodePath: options.sourceNodePath,
+        concurrency: options.concurrency,
+        retryOptions: options.retryOptions
+      });
       return { ok: true, success: true, status: "success", value };
     }
-    const results = [];
-    for (const [index, chunk] of chunks.entries()) {
-      const input = typeof chunk === "string" ? { ssml: chunk } : chunk;
-      const sourceNodePath = input.sourceNodePath;
-      pending(index, "synthesizing");
-      const startedAt = Date.now();
-      try {
-        const result = await client.synthesizeSsml(input.ssml);
-        results.push({
-          ...result,
-          ...input.originalTextRange ? { textRange: { ...input.originalTextRange } } : {},
-          ...sourceNodePath ? {
-            boundaries: result.boundaries?.map((event) => ({
-              ...event,
-              sourceNodePath: [...sourceNodePath]
-            })),
-            visemes: result.visemes?.map((event) => ({ ...event, sourceNodePath: [...sourceNodePath] })),
-            bookmarks: result.bookmarks?.map((event) => ({ ...event, sourceNodePath: [...sourceNodePath] }))
-          } : {}
-        });
-        options.onProgress?.({
-          currentChunk: index + 1,
-          totalChunks: chunks.length,
-          percent: chunks.length === 0 ? 100 : Math.round((index + 1) / chunks.length * 100),
-          chunkIndex: index,
-          originalTextRange: input.originalTextRange,
-          status: "success",
-          durationMs: Date.now() - startedAt
-        });
-      } catch (error) {
-        options.onProgress?.({
-          currentChunk: index,
-          totalChunks: chunks.length,
-          percent: chunks.length === 0 ? 100 : Math.round(index / chunks.length * 100),
-          chunkIndex: index,
-          originalTextRange: input.originalTextRange,
-          status: "failed",
-          durationMs: Date.now() - startedAt,
-          error
-        });
-        throw error;
+    const results = new Array(chunks.length);
+    let completed = 0;
+    let nextIndex = 0;
+    const concurrency = resolveConcurrency2(options.concurrency, chunks.length);
+    const worker = async () => {
+      while (true) {
+        const index = nextIndex++;
+        if (index >= chunks.length) return;
+        const chunk = chunks[index];
+        const input = typeof chunk === "string" ? { ssml: chunk } : chunk;
+        const sourceNodePath = input.sourceNodePath;
+        const originalTextRange = input.originalTextRange;
+        pending(index, "synthesizing");
+        const startedAt = Date.now();
+        try {
+          const result = await retryableSynthesis(
+            () => client.synthesizeSsml(input.ssml, {
+              outputFormat: options.outputFormat,
+              signal: options.signal,
+              timeoutMs: options.timeoutMs,
+              sourceNodePath: input.sourceNodePath ?? options.sourceNodePath
+            }),
+            options.retryOptions,
+            options.signal,
+            (retryAttempt, nextRetryDelayMs) => options.onProgress?.({
+              currentChunk: completed,
+              totalChunks: chunks.length,
+              percent: chunks.length === 0 ? 100 : Math.round(completed / chunks.length * 100),
+              chunkIndex: index,
+              originalTextRange: input.originalTextRange,
+              status: "synthesizing",
+              durationMs: Date.now() - startedAt,
+              retryAttempt,
+              nextRetryDelayMs,
+              isRetrying: true
+            })
+          );
+          results[index] = {
+            ...result,
+            ...input.originalTextRange ? { textRange: { ...input.originalTextRange } } : {},
+            ...sourceNodePath ? {
+              boundaries: result.boundaries?.map((event) => ({
+                ...event,
+                sourceNodePath: [...sourceNodePath],
+                ...event.originalTextRange ? { originalTextRange: { ...event.originalTextRange } } : input.originalTextRange ? { originalTextRange: { ...input.originalTextRange } } : {}
+              })),
+              visemes: result.visemes?.map((event) => ({
+                ...event,
+                sourceNodePath: [...sourceNodePath],
+                ...event.originalTextRange ? { originalTextRange: { ...event.originalTextRange } } : input.originalTextRange ? { originalTextRange: { ...input.originalTextRange } } : {}
+              })),
+              bookmarks: result.bookmarks?.map((event) => ({
+                ...event,
+                sourceNodePath: [...sourceNodePath],
+                ...event.originalTextRange ? { originalTextRange: { ...event.originalTextRange } } : input.originalTextRange ? { originalTextRange: { ...input.originalTextRange } } : {}
+              }))
+            } : {},
+            ...originalTextRange ? {
+              boundaries: result.boundaries?.map((event) => ({
+                ...event,
+                originalTextRange: event.originalTextRange ? { ...event.originalTextRange } : { ...originalTextRange }
+              })),
+              wordBoundary: result.wordBoundary?.map((event) => ({
+                ...event,
+                originalTextRange: event.originalTextRange ? { ...event.originalTextRange } : { ...originalTextRange }
+              })),
+              wordBoundaries: result.wordBoundaries?.map((event) => ({
+                ...event,
+                originalTextRange: event.originalTextRange ? { ...event.originalTextRange } : { ...originalTextRange }
+              })),
+              visemes: result.visemes?.map((event) => ({
+                ...event,
+                originalTextRange: event.originalTextRange ? { ...event.originalTextRange } : { ...originalTextRange }
+              })),
+              bookmarks: result.bookmarks?.map((event) => ({
+                ...event,
+                originalTextRange: event.originalTextRange ? { ...event.originalTextRange } : { ...originalTextRange }
+              }))
+            } : {}
+          };
+          completed += 1;
+          options.onProgress?.({
+            currentChunk: completed,
+            totalChunks: chunks.length,
+            percent: chunks.length === 0 ? 100 : Math.round(completed / chunks.length * 100),
+            chunkIndex: index,
+            originalTextRange: input.originalTextRange,
+            status: "success",
+            durationMs: Date.now() - startedAt
+          });
+        } catch (error) {
+          options.onProgress?.({
+            currentChunk: completed,
+            totalChunks: chunks.length,
+            percent: chunks.length === 0 ? 100 : Math.round(index / chunks.length * 100),
+            chunkIndex: index,
+            originalTextRange: input.originalTextRange,
+            status: "failed",
+            durationMs: Date.now() - startedAt,
+            error
+          });
+          throw error;
+        }
       }
-    }
+    };
+    await Promise.all(Array.from({ length: Math.min(concurrency, chunks.length) }, () => worker()));
+    const orderedResults = results.filter((result) => result !== void 0);
     return {
       ok: true,
       success: true,
       status: "success",
-      value: mergeSynthesisResults(results, options.outputFormat)
+      value: mergeSynthesisResults(orderedResults, {
+        format: options.outputFormat ?? "audio-16khz-128kbitrate-mono-mp3",
+        signal: options.signal
+      })
     };
   } catch (error) {
-    const azureError = error instanceof AzureTtsError ? error : createSpeechSdkError(error);
-    return { ok: false, success: false, status: "azure-api-error", error: azureError };
+    const synthesisError = toSynthesisError(error);
+    return failure(synthesisError);
   }
 }
+function withValidationSignal(options, signal) {
+  if (!signal) return options;
+  return {
+    ...options,
+    urlValidatorSignal: signal,
+    urlValidation: { ...options.urlValidation ?? {}, signal }
+  };
+}
 
 // packages/azure-tts-client/src/client.ts
 var ENDPOINT_TEMPLATE = "https://{region}.tts.speech.microsoft.com/cognitiveservices/v1";
@@ -4707,11 +5613,21 @@ var AzureTtsClient = class {
     const config = { endpoint, region, subscriptionKey, outputFormat, signal, timeoutMs };
     return synthesizeSpeech(ssml, config);
   }
-  async synthesizeSsml(ssml) {
+  async synthesizeSsml(ssml, options = {}) {
     const { region, subscriptionKey, outputFormat, signal, timeoutMs } = __privateGet(this, _options);
     const endpoint = __privateGet(this, _options).endpoint?.trim() || ENDPOINT_TEMPLATE.replace("{region}", region);
     __privateGet(this, _options).logger?.debug?.("Using Azure TTS endpoint:", endpoint);
-    return synthesizeSsml(ssml, { endpoint, region, subscriptionKey, outputFormat, signal, timeoutMs });
+    return synthesizeSsml(ssml, {
+      endpoint,
+      region,
+      subscriptionKey,
+      outputFormat: options.outputFormat ?? outputFormat,
+      signal: options.signal ?? signal,
+      timeoutMs: options.timeoutMs ?? timeoutMs,
+      sourceNodePath: options.sourceNodePath,
+      sourceTextSegments: options.sourceTextSegments,
+      sourceMarkers: options.sourceMarkers
+    });
   }
   async synthesizeChunks(chunks, options = {}) {
     const { region, subscriptionKey, outputFormat, signal, timeoutMs } = __privateGet(this, _options);
@@ -4720,10 +5636,13 @@ var AzureTtsClient = class {
       endpoint,
       region,
       subscriptionKey,
-      outputFormat,
-      signal,
-      timeoutMs,
-      onProgress: options.onProgress ?? __privateGet(this, _options).onProgress
+      outputFormat: options.outputFormat ?? outputFormat,
+      signal: options.signal ?? signal,
+      timeoutMs: options.timeoutMs ?? timeoutMs,
+      sourceNodePath: options.sourceNodePath,
+      onProgress: options.onProgress ?? __privateGet(this, _options).onProgress,
+      concurrency: options.concurrency ?? __privateGet(this, _options).concurrency,
+      retryOptions: options.retryOptions ?? __privateGet(this, _options).retryOptions
     });
   }
   async synthesizeSsmlSafe(ssml, options = {}) {
@@ -4733,7 +5652,11 @@ var AzureTtsClient = class {
     return synthesizeSsmlChunksSafe(this, chunks, {
       ...options,
       outputFormat: options.outputFormat ?? __privateGet(this, _options).outputFormat,
-      onProgress: options.onProgress ?? __privateGet(this, _options).onProgress
+      signal: options.signal ?? __privateGet(this, _options).signal,
+      timeoutMs: options.timeoutMs ?? __privateGet(this, _options).timeoutMs,
+      onProgress: options.onProgress ?? __privateGet(this, _options).onProgress,
+      concurrency: options.concurrency ?? __privateGet(this, _options).concurrency,
+      retryOptions: options.retryOptions ?? __privateGet(this, _options).retryOptions
     });
   }
   async synthesizeSsmlChunksSafe(chunks, options = {}) {
@@ -4830,10 +5753,15 @@ async function fetchAzureVoiceCatalog(options) {
 }
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {
+  AudioFormatMismatchError,
   AzureTtsClient,
   AzureTtsError,
   AzureTtsSdkError,
   ChunkValidationError,
+  DEFAULT_OUTPUT_FORMAT,
+  MergeError,
+  SynthesisCancelledError,
+  SynthesisTimeoutError,
   UnsupportedMergeFormatError,
   areAzureLanguagesEquivalent,
   buildPartialSsml,
@@ -4846,6 +5774,8 @@ async function fetchAzureVoiceCatalog(options) {
   fromPlainTextToSsml,
   getAzureVoiceCatalogMetadata,
   getBuiltInVoiceCatalogMetadata,
+  getSsmlSourceMap,
+  inspectAudioSpecification,
   isValidAzureAudioDuration,
   mapSsmlTextNodes,
   mergeAudioBuffers,
@@ -4853,6 +5783,7 @@ async function fetchAzureVoiceCatalog(options) {
   normalizeAzureLanguage,
   parseSsml,
   resolveMergeAudioFormat,
+  resolveMimeType,
   splitSsmlDocument,
   synthesizeSpeech,
   synthesizeSsml,
@@ -4860,6 +5791,7 @@ async function fetchAzureVoiceCatalog(options) {
   synthesizeSsmlChunksSafe,
   synthesizeSsmlSafe,
   validateAzureSsml,
+  validateAzureSsmlChunks,
   validateSsml,
   validateSsmlStructureIntegrity
 });