ssml-builder-js 2.13.0 → 2.15.0

package/dist/index.js

@@ -40,24 +40,37 @@ __export(src_exports, {
   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,
   buildSsml: () => buildSsml,
+  canMergeAudioFormat: () => canMergeAudioFormat,
+  createAzureUrlValidatorRunner: () => createAzureUrlValidatorRunner,
   extractSsmlText: () => extractSsmlText,
   extractSsmlTranslatableText: () => extractSsmlTranslatableText,
   fetchAzureVoiceCatalog: () => fetchAzureVoiceCatalog,
   fromPlainTextToSsml: () => fromPlainTextToSsml,
   getAzureVoiceCatalogMetadata: () => getAzureVoiceCatalogMetadata,
   getBuiltInVoiceCatalogMetadata: () => getBuiltInVoiceCatalogMetadata,
+  getSsmlSourceMap: () => getSsmlSourceMap,
   isValidAzureAudioDuration: () => isValidAzureAudioDuration,
   mapSsmlTextNodes: () => mapSsmlTextNodes,
+  mergeAudioBuffers: () => mergeAudioBuffers,
   mergeSynthesisResults: () => mergeSynthesisResults,
   normalizeAzureLanguage: () => normalizeAzureLanguage,
   parseSsml: () => parseSsml,
+  resolveMergeAudioFormat: () => resolveMergeAudioFormat,
+  resolveMimeType: () => resolveMimeType,
   splitSsmlDocument: () => splitSsmlDocument,
   synthesizeSpeech: () => synthesizeSpeech,
   synthesizeSsml: () => synthesizeSsml,
   synthesizeSsmlChunks: () => synthesizeSsmlChunks,
+  synthesizeSsmlChunksSafe: () => synthesizeSsmlChunksSafe,
   synthesizeSsmlSafe: () => synthesizeSsmlSafe,
   validateAzureSsml: () => validateAzureSsml,
   validateSsml: () => validateSsml,
@@ -992,6 +1005,236 @@ 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) {
@@ -1095,7 +1338,35 @@ function collectInheritedContext(nodes) {
   nodes.forEach(visit);
   return context;
 }
-function createChunk(document, nodes, chunkIndex, textStart, backgroundAudio, replicateBackgroundAudio) {
+function elementName(node) {
+  return node.type === "custom" || node.type === "element" ? node.name : node.type;
+}
+function findSourceNodePath(nodes, targetOffset) {
+  let textOffset = 0;
+  let firstPath;
+  let foundPath;
+  const visit = (node, path) => {
+    if (typeof node === "string" || node.type === "text") {
+      const text = typeof node === "string" ? node : node.value;
+      if (text && firstPath === void 0) firstPath = [...path];
+      if (text && foundPath === void 0 && targetOffset < textOffset + text.length) foundPath = [...path];
+      textOffset += text.length;
+      return;
+    }
+    node.children?.forEach((child, index) => {
+      const childPath = typeof child === "string" || child.type === "text" ? path : [...path, `${elementName(child)}[${index}]`];
+      visit(child, childPath);
+    });
+  };
+  nodes.forEach((node, index) => {
+    if (!foundPath) {
+      if (typeof node === "string" || node.type === "text") visit(node, ["speak"]);
+      else visit(node, ["speak", `${elementName(node)}[${index}]`]);
+    }
+  });
+  return foundPath ?? firstPath;
+}
+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 = [];
@@ -1110,7 +1381,24 @@ function createChunk(document, nodes, chunkIndex, textStart, backgroundAudio, re
     containedMarks: marks,
     hasBackgroundAudio: chunkNodes.some(
       (node) => typeof node !== "string" && node.type !== "text" && node.type === "mstts:backgroundaudio"
-    )
+    ),
+    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 = {}) {
@@ -1120,11 +1408,12 @@ 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;
@@ -1148,7 +1437,16 @@ 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");
     }
@@ -1162,7 +1460,9 @@ 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;
@@ -1185,172 +1485,23 @@ 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;
-    }
-  }
-  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 elementName(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;
+// 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) {
@@ -1489,7 +1640,7 @@ function extractSsmlTranslatableText(ssml, options = {}) {
         }
         return;
       }
-      const tag = elementName(node);
+      const tag = elementName2(node);
       if (skipTags.has(tag.toLowerCase())) return;
       visit(childrenOf(node), [...ancestors, tag], [...path, String(index)]);
     });
@@ -1535,7 +1686,7 @@ function serializeDocument2(document) {
   const serialize = (node) => {
     if (typeof node === "string") return node.replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;");
     if (node.type === "text") return serialize(node.value);
-    const tag = elementName(node);
+    const tag = elementName2(node);
     const nodeAttributes = elementAttributes(node);
     const serializedAttributes = Object.entries(nodeAttributes).map(([name, value]) => ` ${name}="${serialize(value).replace(/"/g, "&quot;")}"`).join("");
     const children = childrenOf(node).map(serialize).join("");
@@ -1549,7 +1700,7 @@ function flatten(document) {
     nodes.forEach((node, index) => {
       if (typeof node === "string" || node.type === "text") return;
       const currentPath = `${path}/${index}`;
-      result.push({ name: elementName(node), attributes: elementAttributes(node), path: currentPath });
+      result.push({ name: elementName2(node), attributes: elementAttributes(node), path: currentPath });
       visit(childrenOf(node), currentPath);
     });
   };
@@ -1747,6 +1898,86 @@ var AZURE_VOICE_DEFINITIONS = [
 ];
 
 // packages/ssml-core/src/azureValidation.ts
+function createAzureUrlValidatorRunner(validator, options = {}) {
+  if (typeof validator !== "function") throw new TypeError("A URL validator function is required.");
+  const concurrency = options.concurrency === void 0 ? Infinity : Number.isFinite(options.concurrency) ? Math.max(1, Math.floor(options.concurrency)) : Infinity;
+  const cache = options.cache ?? /* @__PURE__ */ new Map();
+  const inFlight = /* @__PURE__ */ new Map();
+  const waiters = [];
+  let active = 0;
+  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, 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, 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(key);
+    if (existing) return existing;
+    const promise = (async () => {
+      await acquire(signal);
+      try {
+        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."));
+          signal.addEventListener("abort", abortHandler, { once: true });
+          if (options.timeoutMs !== void 0 && options.timeoutMs > 0) {
+            timer = setTimeout(
+              () => reject(new Error(`URL validation timed out after ${options.timeoutMs} ms.`)),
+              options.timeoutMs
+            );
+          }
+        });
+        try {
+          const result = await (timer || abortHandler ? Promise.race([validation, cancellation]) : validation);
+          cache.set(key, result);
+          return result;
+        } finally {
+          if (timer) clearTimeout(timer);
+          if (abortHandler) signal.removeEventListener("abort", abortHandler);
+        }
+      } finally {
+        release();
+      }
+    })();
+    inFlight.set(key, promise);
+    try {
+      return await promise;
+    } finally {
+      inFlight.delete(key);
+    }
+  };
+  return (url, context, signal) => check(url, context, signal ?? configuredSignal);
+}
 var ALLOWED_BREAK_STRENGTHS = /* @__PURE__ */ new Set(["none", "x-weak", "weak", "medium", "strong", "x-strong"]);
 var ALLOWED_SAY_AS = /* @__PURE__ */ new Set([
   "characters",
@@ -2041,23 +2272,23 @@ function validateVoiceFeatureMatrix(token, source, diagnostics, voiceName, defin
     );
   }
 }
-function validateAudioSource(token, source, diagnostics, options, elementName2) {
+function validateAudioSource(token, source, diagnostics, options, elementName3) {
   const src = attr(token, "src");
   if (!src) {
-    addDiagnostic(diagnostics, source, token.start, `<${elementName2}> requires a "src" attribute.`);
+    addDiagnostic(diagnostics, source, token.start, `<${elementName3}> requires a "src" attribute.`);
     return;
   }
   let parsed;
   try {
     parsed = new URL(src);
   } catch {
-    addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> must be an absolute HTTP(S) URL.`);
+    addDiagnostic(diagnostics, source, token.start, `<${elementName3} src> must be an absolute HTTP(S) URL.`);
     return;
   }
   if (parsed.username || parsed.password)
-    addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> must not contain URL credentials.`);
+    addDiagnostic(diagnostics, source, token.start, `<${elementName3} src> must not contain URL credentials.`);
   if (parsed.protocol !== "https:" && !(options.allowHttpAudio && parsed.protocol === "http:"))
-    addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> must use HTTPS.`);
+    addDiagnostic(diagnostics, source, token.start, `<${elementName3} src> must use HTTPS.`);
   const isAllowedOrigin = options.allowedAudioOrigins?.some((allowedOrigin) => {
     try {
       const configured = new URL(allowedOrigin);
@@ -2069,13 +2300,13 @@ function validateAudioSource(token, source, diagnostics, options, elementName2)
     }
   }) ?? false;
   if (options.allowedAudioOrigins && !isAllowedOrigin)
-    addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> origin "${parsed.origin}" is not allowed.`);
+    addDiagnostic(diagnostics, source, token.start, `<${elementName3} src> origin "${parsed.origin}" is not allowed.`);
   else if (!isAllowedOrigin && !options.allowExternalAudio)
     addDiagnostic(
       diagnostics,
       source,
       token.start,
-      `<${elementName2} src> external origin "${parsed.origin}" is blocked by default; set allowExternalAudio to true or provide allowedAudioOrigins.`
+      `<${elementName3} src> external origin "${parsed.origin}" is blocked by default; set allowExternalAudio to true or provide allowedAudioOrigins.`
     );
 }
 function validateElement(token, source, diagnostics, voiceName, options, voiceCatalog) {
@@ -2393,6 +2624,15 @@ 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, {
+    ...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);
@@ -2402,7 +2642,11 @@ function validateAzureSsml(ssml, options = {}) {
   const checks = tokens.flatMap(
     (token) => urlAttributes(token).map(async ({ attribute, value }) => {
       try {
-        const result = await validator(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;
@@ -2448,6 +2692,7 @@ var getBuiltInVoiceCatalogMetadata = getAzureVoiceCatalogMetadata;
 var AzureTtsError = class extends Error {
   constructor(status, statusText, responseBody, requestId) {
     super(`Azure TTS request failed: ${status} ${statusText}`);
+    this.kind = "azure-api-error";
     this.name = "AzureTtsError";
     this.status = status;
     this.statusText = statusText;
@@ -2463,6 +2708,44 @@ var AzureTtsSdkError = class extends AzureTtsError {
     this.errorDetails = errorDetails;
   }
 };
+var SynthesisCancelledError = class extends Error {
+  constructor(message = "Speech synthesis was cancelled.") {
+    super(message);
+    this.kind = "cancelled";
+    this.name = "SynthesisCancelledError";
+  }
+};
+var SynthesisTimeoutError = class extends Error {
+  constructor(message) {
+    super(message);
+    this.kind = "timeout";
+    this.name = "SynthesisTimeoutError";
+  }
+};
+var MergeError = class extends Error {
+  constructor(message, cause) {
+    super(message);
+    this.kind = "merge-error";
+    this.name = "MergeError";
+    this.cause = cause;
+  }
+};
+var UnsupportedMergeFormatError = class extends Error {
+  constructor(format) {
+    super(`Audio format "${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 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);
@@ -2471,263 +2754,9 @@ 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
-};
-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 closeSpeechResources(speechConfig, synthesizer) {
-  try {
-    synthesizer.close();
-  } catch {
-  }
-  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 } } : {},
-        ...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;
-  for (const [index, chunk] of chunks.entries()) {
-    const input = typeof chunk === "string" ? { ssml: chunk } : chunk;
-    const result = await synthesizeSsml(input.ssml, {
-      ...config,
-      ...input.originalTextRange ? { sourceTextRange: input.originalTextRange } : {},
-      onProgress: void 0
-    });
-    results.push(result);
-    config.onProgress?.({
-      currentChunk: index + 1,
-      totalChunks,
-      percent: totalChunks === 0 ? 100 : Math.round((index + 1) / totalChunks * 100)
-    });
-  }
-  return mergeSynthesisResults(results);
-}
-function mergeSynthesisResults(results) {
-  const audioLength = results.reduce((total, result) => total + result.audioData.byteLength, 0);
-  const audioData = new Uint8Array(audioLength);
-  const boundaries = [];
-  const visemes = [];
-  const bookmarks = [];
-  let byteOffset = 0;
-  let durationOffset = 0;
-  for (const result of results) {
-    audioData.set(new Uint8Array(result.audioData), byteOffset);
-    byteOffset += result.audioData.byteLength;
-    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 requestId = boundary.requestId ?? result.requestId;
-      boundaries.push({
-        ...boundary,
-        audioOffsetMs: boundary.audioOffsetMs + durationOffset,
-        ...textRange ? { textRange: { ...textRange } } : {},
-        ...requestId ? { requestId } : {}
-      });
-    }
-    for (const viseme of result.visemes ?? []) {
-      const textRange = viseme.textRange ?? result.textRange;
-      const requestId = viseme.requestId ?? result.requestId;
-      visemes.push({
-        ...viseme,
-        audioOffsetMs: viseme.audioOffsetMs + durationOffset,
-        ...textRange ? { textRange: { ...textRange } } : {},
-        ...requestId ? { requestId } : {}
-      });
-    }
-    for (const bookmark of result.bookmarks ?? []) {
-      const textRange = bookmark.textRange ?? result.textRange;
-      const requestId = bookmark.requestId ?? result.requestId;
-      bookmarks.push({
-        ...bookmark,
-        audioOffsetMs: bookmark.audioOffsetMs + durationOffset,
-        ...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);
+// 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));
@@ -3376,6 +3405,130 @@ 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"] },
@@ -3496,6 +3649,86 @@ var AZURE_VOICE_DEFINITIONS2 = [
   },
   { name: "zh-TW-HsiaoChenNeural", locale: "zh-TW" }
 ];
+function createAzureUrlValidatorRunner2(validator, options = {}) {
+  if (typeof validator !== "function") throw new TypeError("A URL validator function is required.");
+  const concurrency = options.concurrency === void 0 ? Infinity : Number.isFinite(options.concurrency) ? Math.max(1, Math.floor(options.concurrency)) : Infinity;
+  const cache = options.cache ?? /* @__PURE__ */ new Map();
+  const inFlight = /* @__PURE__ */ new Map();
+  const waiters = [];
+  let active = 0;
+  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, 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, 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(key);
+    if (existing) return existing;
+    const promise = (async () => {
+      await acquire(signal);
+      try {
+        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."));
+          signal.addEventListener("abort", abortHandler, { once: true });
+          if (options.timeoutMs !== void 0 && options.timeoutMs > 0) {
+            timer = setTimeout(
+              () => reject(new Error(`URL validation timed out after ${options.timeoutMs} ms.`)),
+              options.timeoutMs
+            );
+          }
+        });
+        try {
+          const result = await (timer || abortHandler ? Promise.race([validation, cancellation]) : validation);
+          cache.set(key, result);
+          return result;
+        } finally {
+          if (timer) clearTimeout(timer);
+          if (abortHandler) signal.removeEventListener("abort", abortHandler);
+        }
+      } finally {
+        release();
+      }
+    })();
+    inFlight.set(key, promise);
+    try {
+      return await promise;
+    } finally {
+      inFlight.delete(key);
+    }
+  };
+  return (url, context, signal) => check(url, context, signal ?? configuredSignal);
+}
 var ALLOWED_BREAK_STRENGTHS2 = /* @__PURE__ */ new Set(["none", "x-weak", "weak", "medium", "strong", "x-strong"]);
 var ALLOWED_SAY_AS2 = /* @__PURE__ */ new Set([
   "characters",
@@ -3780,23 +4013,23 @@ function validateVoiceFeatureMatrix2(token, source, diagnostics, voiceName, defi
     );
   }
 }
-function validateAudioSource2(token, source, diagnostics, options, elementName2) {
+function validateAudioSource2(token, source, diagnostics, options, elementName3) {
   const src = attr2(token, "src");
   if (!src) {
-    addDiagnostic2(diagnostics, source, token.start, `<${elementName2}> requires a "src" attribute.`);
+    addDiagnostic2(diagnostics, source, token.start, `<${elementName3}> requires a "src" attribute.`);
     return;
   }
   let parsed;
   try {
     parsed = new URL(src);
   } catch {
-    addDiagnostic2(diagnostics, source, token.start, `<${elementName2} src> must be an absolute HTTP(S) URL.`);
+    addDiagnostic2(diagnostics, source, token.start, `<${elementName3} src> must be an absolute HTTP(S) URL.`);
     return;
   }
   if (parsed.username || parsed.password)
-    addDiagnostic2(diagnostics, source, token.start, `<${elementName2} src> must not contain URL credentials.`);
+    addDiagnostic2(diagnostics, source, token.start, `<${elementName3} src> must not contain URL credentials.`);
   if (parsed.protocol !== "https:" && !(options.allowHttpAudio && parsed.protocol === "http:"))
-    addDiagnostic2(diagnostics, source, token.start, `<${elementName2} src> must use HTTPS.`);
+    addDiagnostic2(diagnostics, source, token.start, `<${elementName3} src> must use HTTPS.`);
   const isAllowedOrigin = options.allowedAudioOrigins?.some((allowedOrigin) => {
     try {
       const configured = new URL(allowedOrigin);
@@ -3808,13 +4041,13 @@ function validateAudioSource2(token, source, diagnostics, options, elementName2)
     }
   }) ?? false;
   if (options.allowedAudioOrigins && !isAllowedOrigin)
-    addDiagnostic2(diagnostics, source, token.start, `<${elementName2} src> origin "${parsed.origin}" is not allowed.`);
+    addDiagnostic2(diagnostics, source, token.start, `<${elementName3} src> origin "${parsed.origin}" is not allowed.`);
   else if (!isAllowedOrigin && !options.allowExternalAudio)
     addDiagnostic2(
       diagnostics,
       source,
       token.start,
-      `<${elementName2} src> external origin "${parsed.origin}" is blocked by default; set allowExternalAudio to true or provide allowedAudioOrigins.`
+      `<${elementName3} src> external origin "${parsed.origin}" is blocked by default; set allowExternalAudio to true or provide allowedAudioOrigins.`
     );
 }
 function validateElement2(token, source, diagnostics, voiceName, options, voiceCatalog) {
@@ -4132,6 +4365,15 @@ 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 } : {}
+  });
+  const validationSignal = options.urlValidatorSignal ?? options.urlValidation?.signal ?? new AbortController().signal;
   let tokens;
   try {
     tokens = tokenizeElements2(ssml);
@@ -4141,7 +4383,11 @@ function validateAzureSsml2(ssml, options = {}) {
   const checks = tokens.flatMap(
     (token) => urlAttributes2(token).map(async ({ attribute, value }) => {
       try {
-        const result = await validator(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;
@@ -4172,51 +4418,793 @@ var AZURE_VOICE_CATALOG_METADATA2 = {
   voiceCount: AZURE_VOICE_DEFINITIONS2.length
 };
 
-// packages/azure-tts-client/src/safe.ts
-async function synthesizeSsmlSafe(client, ssml, options = {}) {
-  const validationOptions = options.validation ?? options;
-  const diagnostics = await Promise.resolve(validateAzureSsml2(ssml, validationOptions));
-  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
-      }
-    };
-  }
-  try {
-    return { ok: true, success: true, status: "success", value: await client.synthesizeSsml(ssml) };
-  } catch (error) {
-    const azureError = error instanceof AzureTtsError ? error : createSpeechSdkError(error);
-    return { ok: false, success: false, status: "azure-api-error", error: azureError };
-  }
-}
-
-// packages/azure-tts-client/src/client.ts
-var ENDPOINT_TEMPLATE = "https://{region}.tts.speech.microsoft.com/cognitiveservices/v1";
-var _options;
-var AzureTtsClient = class {
-  constructor(options) {
-    __privateAdd(this, _options);
-    __privateSet(this, _options, options);
-  }
-  async synthesize(ssml) {
-    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);
+// 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 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 (!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(?:-|$)/i.test(format);
+}
+function resolveMergeAudioFormat(format) {
+  if (isWavFormat(format)) return "wav";
+  if (isMp3Format(format)) return "mp3";
+  if (isRawFormat(format)) return "raw";
+  return void 0;
+}
+function canMergeAudioFormat(format) {
+  return resolveMergeAudioFormat(format) !== void 0;
+}
+function mergeAudioBuffers(buffers, options) {
+  const format = typeof options === "string" ? options : options?.format;
+  if (!format) throw new UnsupportedMergeFormatError("");
+  try {
+    if (isWavFormat(format)) return mergeWavBuffers(buffers);
+    if (isMp3Format(format)) {
+      const parts = buffers.map(stripMp3Tags);
+      const output = new Uint8Array(parts.reduce((total, part) => total + part.byteLength, 0));
+      let offset = 0;
+      for (const part of parts) {
+        output.set(part, offset);
+        offset += part.byteLength;
+      }
+      return output.buffer;
+    }
+    if (isRawFormat(format)) {
+      const output = new Uint8Array(buffers.reduce((total, buffer) => total + buffer.byteLength, 0));
+      let offset = 0;
+      for (const buffer of buffers) {
+        output.set(new Uint8Array(buffer), offset);
+        offset += buffer.byteLength;
+      }
+      return output.buffer;
+    }
+    throw new UnsupportedMergeFormatError(format);
+  } catch (error) {
+    if (error instanceof UnsupportedMergeFormatError || error instanceof MergeError) throw error;
+    throw new MergeError(`Audio buffers could not be merged for format "${format}".`, error);
+  }
+}
+function closeSpeechResources(speechConfig, synthesizer) {
+  try {
+    synthesizer.close();
+  } catch {
+  }
+  try {
+    speechConfig.close();
+  } catch {
+  }
+}
+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 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 }
+        };
+      }
+      if (sourceSegments.length === 0 && !config.sourceTextRange && !config.sourceNodePath) return {};
+      const value = text ?? "";
+      let localStart = Number.isFinite(offsetHint) && (offsetHint ?? 0) >= 0 ? offsetHint : -1;
+      if (value && localStart >= 0 && sourceText.slice(localStart, localStart + value.length) !== value)
+        localStart = -1;
+      if (localStart < 0 || localStart > sourceText.length) {
+        localStart = value ? sourceText.indexOf(value, sourceEventCursor) : sourceEventCursor;
+        if (localStart < 0) localStart = value ? sourceText.indexOf(value) : sourceEventCursor;
+      }
+      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] } : {}
+      };
+    };
+    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) => ({
+        ...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 } : {}
+      });
+      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(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);
+    }
+  });
+}
+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 ?? config.sourceNodePath ? { sourceNodePath: [...input.sourceNodePath ?? config.sourceNodePath ?? []] } : {},
+        ...input.sourceTextSegments ? { sourceTextSegments: input.sourceTextSegments } : {},
+        ...input.sourceMarkers ? { sourceMarkers: input.sourceMarkers } : {},
+        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, {
+    format: config.outputFormat ?? "audio-16khz-128kbitrate-mono-mp3"
+  });
+}
+function createMergedResult(results, audioData, format) {
+  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 } : {}
+      });
+    }
+    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 } : {}
+      });
+    }
+    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 } : {}
+      });
+    }
+    durationOffset += Math.max(0, result.durationMs);
+  }
+  return {
+    audioData,
+    durationMs: durationOffset,
+    mimeType: resolveMimeType(format),
+    ...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);
+  if (resolvedOptions.customMerger) {
+    return Promise.resolve().then(() => resolvedOptions.customMerger?.(buffers, format)).then((merged) => {
+      if (!merged) throw new MergeError("The custom audio merger returned no audio buffer.");
+      return createMergedResult(results, merged, format);
+    }).catch((error) => {
+      if (error instanceof UnsupportedMergeFormatError || error instanceof MergeError) throw error;
+      throw new MergeError(`Custom audio merger failed for format "${format}".`, error);
+    });
+  }
+  try {
+    return createMergedResult(results, mergeAudioBuffers(buffers, { format }), format);
+  } catch (error) {
+    if (error instanceof UnsupportedMergeFormatError || 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;
+}
+
+// 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 = "validation-error";
+    this.name = "ChunkValidationError";
+    this.chunkIndex = chunkIndex;
+    this.diagnostics = diagnostics;
+  }
+};
+function failure(error) {
+  return { ok: false, success: false, status: error.kind, error };
+}
+async function synthesizeSsmlSafe(client, ssml, options = {}) {
+  const validationOptions = withValidationSignal(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 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, { signal: options.signal })
+    };
+  } catch (error) {
+    const synthesisError = toSynthesisError(error);
+    return failure(synthesisError);
+  }
+}
+async function synthesizeSsmlChunksSafe(client, chunks, options = {}) {
+  const validationOptions = withValidationSignal(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,
+      totalChunks: chunks.length,
+      percent: chunks.length === 0 ? 100 : Math.round((status === "success" ? index + 1 : index) / chunks.length * 100),
+      chunkIndex: index,
+      originalTextRange: typeof chunks[index] === "string" ? void 0 : chunks[index]?.originalTextRange,
+      status,
+      durationMs: 0,
+      ...error ? { error } : {}
+    });
+  };
+  chunks.forEach((_chunk, index) => {
+    pending(index, "pending");
+  });
+  const validations = await Promise.all(
+    chunks.map(async (chunk) => {
+      const ssml = typeof chunk === "string" ? chunk : chunk.ssml;
+      const sourceNodePath = typeof chunk === "string" ? options.sourceNodePath : chunk.sourceNodePath ?? options.sourceNodePath;
+      const diagnostics = await Promise.resolve(
+        validateAzureSsml2(ssml, { ...validationOptions, ...sourceNodePath ? { sourceNodePath } : {} })
+      );
+      return diagnostics.filter((diagnostic) => diagnostic.severity === "error");
+    })
+  );
+  const firstInvalidIndex = validations.findIndex((diagnostics) => diagnostics.length > 0);
+  if (firstInvalidIndex >= 0) {
+    const error = new ChunkValidationError(firstInvalidIndex, validations[firstInvalidIndex] ?? []);
+    pending(firstInvalidIndex, "failed", error);
+    return failure(error);
+  }
+  try {
+    if (client.synthesizeChunks) {
+      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
+      });
+      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;
+      const originalTextRange = input.originalTextRange;
+      pending(index, "synthesizing");
+      const startedAt = Date.now();
+      try {
+        const result = await client.synthesizeSsml(input.ssml, {
+          outputFormat: options.outputFormat,
+          signal: options.signal,
+          timeoutMs: options.timeoutMs,
+          sourceNodePath: input.sourceNodePath ?? options.sourceNodePath
+        });
+        results.push({
+          ...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 }
+            }))
+          } : {}
+        });
+        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;
+      }
+    }
+    return {
+      ok: true,
+      success: true,
+      status: "success",
+      value: mergeSynthesisResults(results, {
+        format: options.outputFormat ?? "audio-16khz-128kbitrate-mono-mp3"
+      })
+    };
+  } catch (error) {
+    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";
+var _options;
+var AzureTtsClient = class {
+  constructor(options) {
+    __privateAdd(this, _options);
+    __privateSet(this, _options, options);
+  }
+  async synthesize(ssml) {
+    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);
     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);
@@ -4225,15 +5213,28 @@ var AzureTtsClient = class {
       endpoint,
       region,
       subscriptionKey,
-      outputFormat,
-      signal,
-      timeoutMs,
+      outputFormat: options.outputFormat ?? outputFormat,
+      signal: options.signal ?? signal,
+      timeoutMs: options.timeoutMs ?? timeoutMs,
+      sourceNodePath: options.sourceNodePath,
       onProgress: options.onProgress ?? __privateGet(this, _options).onProgress
     });
   }
   async synthesizeSsmlSafe(ssml, options = {}) {
     return synthesizeSsmlSafe(this, ssml, options);
   }
+  async synthesizeChunksSafe(chunks, options = {}) {
+    return synthesizeSsmlChunksSafe(this, chunks, {
+      ...options,
+      outputFormat: options.outputFormat ?? __privateGet(this, _options).outputFormat,
+      signal: options.signal ?? __privateGet(this, _options).signal,
+      timeoutMs: options.timeoutMs ?? __privateGet(this, _options).timeoutMs,
+      onProgress: options.onProgress ?? __privateGet(this, _options).onProgress
+    });
+  }
+  async synthesizeSsmlChunksSafe(chunks, options = {}) {
+    return this.synthesizeChunksSafe(chunks, options);
+  }
 };
 _options = new WeakMap();
 
@@ -4289,6 +5290,9 @@ async function fetchAzureVoiceCatalog(options) {
       const secondaryLocales = stringList(record.SecondaryLocaleList);
       const styles = stringList(record.StyleList);
       const status = normalizeStatus(record.Status);
+      const supportedTags = stringList(record.SupportedTags);
+      const unsupportedTags = stringList(record.UnsupportedTags);
+      const models = stringList(record.Models);
       const merged = {
         name: existing?.name ?? name,
         locale: existing?.locale ?? locale,
@@ -4298,6 +5302,12 @@ async function fetchAzureVoiceCatalog(options) {
       if (mergedSecondaryLocales.length > 0) merged.secondaryLocales = mergedSecondaryLocales;
       const mergedStyles = [.../* @__PURE__ */ new Set([...existing?.styles ?? [], ...styles])];
       if (mergedStyles.length > 0) merged.styles = mergedStyles;
+      const mergedSupportedTags = [.../* @__PURE__ */ new Set([...existing?.supportedTags ?? [], ...supportedTags])];
+      if (mergedSupportedTags.length > 0) merged.supportedTags = mergedSupportedTags;
+      const mergedUnsupportedTags = [.../* @__PURE__ */ new Set([...existing?.unsupportedTags ?? [], ...unsupportedTags])];
+      if (mergedUnsupportedTags.length > 0) merged.unsupportedTags = mergedUnsupportedTags;
+      const mergedModels = [.../* @__PURE__ */ new Set([...existing?.models ?? [], ...models])];
+      if (mergedModels.length > 0) merged.models = mergedModels;
       if (status) merged.status = status;
       else if (existing?.status) merged.status = existing.status;
       voices.set(key, merged);
@@ -4319,24 +5329,37 @@ async function fetchAzureVoiceCatalog(options) {
   AzureTtsClient,
   AzureTtsError,
   AzureTtsSdkError,
+  ChunkValidationError,
+  DEFAULT_OUTPUT_FORMAT,
+  MergeError,
+  SynthesisCancelledError,
+  SynthesisTimeoutError,
+  UnsupportedMergeFormatError,
   areAzureLanguagesEquivalent,
   buildPartialSsml,
   buildSsml,
+  canMergeAudioFormat,
+  createAzureUrlValidatorRunner,
   extractSsmlText,
   extractSsmlTranslatableText,
   fetchAzureVoiceCatalog,
   fromPlainTextToSsml,
   getAzureVoiceCatalogMetadata,
   getBuiltInVoiceCatalogMetadata,
+  getSsmlSourceMap,
   isValidAzureAudioDuration,
   mapSsmlTextNodes,
+  mergeAudioBuffers,
   mergeSynthesisResults,
   normalizeAzureLanguage,
   parseSsml,
+  resolveMergeAudioFormat,
+  resolveMimeType,
   splitSsmlDocument,
   synthesizeSpeech,
   synthesizeSsml,
   synthesizeSsmlChunks,
+  synthesizeSsmlChunksSafe,
   synthesizeSsmlSafe,
   validateAzureSsml,
   validateSsml,