ssml-builder-js 2.11.0 → 2.12.0

package/dist/elements.mjs

@@ -36,7 +36,7 @@ import {
   resolveExpressAsStyles,
   updateEditableText,
   validateAzureSsml
-} from "./chunk-FWWMWI2C.mjs";
+} from "./chunk-FHDFRM2F.mjs";
 import "./chunk-6S5ODO6A.mjs";
 
 // packages/ssml-editor-react/src/ssmlInsertions.ts

package/dist/index.d.mts

@@ -1,4 +1,4 @@
-export { AudioElement, AzureDiagnosticCode, AzureLanguageNormalizationOptions, AzureSsmlValidationOptions, AzureValidationOptions, AzureVoiceCatalogMetadata, AzureVoiceDefinition, AzureVoiceMetadata, BookmarkElement, BreakElement, BuildPartialSsmlOptions, CustomElement, EmphasisElement, ExpressAsElement, ExtractSsmlTranslatableTextOptions, FromPlainTextToSsmlOptions, LangElement, LexiconElement, MapSsmlTextNodesOptions, MarkElement, MsttsAudioDurationElement, MsttsEmbeddingElement, MsttsSilenceElement, MsttsTtsEmbeddingElement, MsttsVisemeElement, MsttsVoiceConversionElement, NamedElement, ParagraphElement, PhonemeElement, ProsodyElement, SayAsElement, SentenceElement, SsmlAttributeValue, SsmlAttributes, SsmlBackgroundAudioNode, SsmlBreakElement, SsmlDiagnostic, SsmlDiagnosticSeverity, SsmlDiagnosticSource, SsmlDialogNode, SsmlDocument, SsmlElement, SsmlElementBase, SsmlEmbeddingNode, SsmlExpressAsElement, SsmlNode, SsmlPartialContext, SsmlPartialProsody, SsmlPartialVoice, SsmlPhonemeElement, SsmlProsodyElement, SsmlSayAsElement, SsmlStructureIntegrityResult, SsmlStructureMismatch, SsmlText, SsmlTextNodeContext, SsmlTtsEmbeddingNode, SsmlTurnNode, SsmlValidationError, SsmlVoiceConversionNode, SsmlVoiceElement, SubElement, VoiceElement, WordElement, areAzureLanguagesEquivalent, buildPartialSsml, buildSsml, extractSsmlText, extractSsmlTranslatableText, fromPlainTextToSsml, getAzureVoiceCatalogMetadata, getBuiltInVoiceCatalogMetadata, isValidAzureAudioDuration, mapSsmlTextNodes, normalizeAzureLanguage, parseSsml, validateAzureSsml, validateSsml, validateSsmlStructureIntegrity } from './core.mjs';
+export { AudioElement, AzureDiagnosticCode, AzureLanguageNormalizationOptions, AzureSsmlValidationOptions, AzureValidationOptions, AzureVoiceCatalogMetadata, AzureVoiceDefinition, AzureVoiceMetadata, BookmarkElement, BreakElement, BuildPartialSsmlOptions, CustomElement, EmphasisElement, ExpressAsElement, ExtractSsmlTranslatableTextOptions, FromPlainTextToSsmlOptions, LangElement, LexiconElement, MapSsmlTextNodesOptions, MarkElement, MsttsAudioDurationElement, MsttsEmbeddingElement, MsttsSilenceElement, MsttsTtsEmbeddingElement, MsttsVisemeElement, MsttsVoiceConversionElement, NamedElement, ParagraphElement, PhonemeElement, ProsodyElement, SayAsElement, SentenceElement, SsmlAttributeValue, SsmlAttributes, SsmlBackgroundAudioNode, SsmlBreakElement, SsmlDiagnostic, SsmlDiagnosticSeverity, SsmlDiagnosticSource, SsmlDialogNode, SsmlDocument, SsmlElement, SsmlElementBase, SsmlEmbeddingNode, SsmlExpressAsElement, SsmlNode, SsmlPartialContext, SsmlPartialProsody, SsmlPartialVoice, SsmlPhonemeElement, SsmlProsodyElement, SsmlSayAsElement, SsmlStructureIntegrityResult, SsmlStructureMismatch, SsmlText, SsmlTextNodeContext, SsmlTtsEmbeddingNode, SsmlTurnNode, SsmlValidationError, SsmlVoiceConversionNode, SsmlVoiceElement, SubElement, VoiceElement, WordElement, areAzureLanguagesEquivalent, buildPartialSsml, buildSsml, extractSsmlText, extractSsmlTranslatableText, fromPlainTextToSsml, getAzureVoiceCatalogMetadata, getBuiltInVoiceCatalogMetadata, isValidAzureAudioDuration, mapSsmlTextNodes, normalizeAzureLanguage, parseSsml, splitSsmlDocument, validateAzureSsml, validateSsml, validateSsmlStructureIntegrity } from './core.mjs';
 
 interface TtsConfig {
     signal?: AbortSignal;
@@ -8,6 +8,31 @@ interface TtsConfig {
     region: string;
     outputFormat?: string;
 }
+interface SsmlSynthesisBoundary {
+    text: string;
+    audioOffsetMs: number;
+    durationMs: number;
+}
+interface SsmlSynthesisViseme {
+    visemeId: number;
+    audioOffsetMs: number;
+}
+interface SsmlSynthesisBookmark {
+    name: string;
+    audioOffsetMs: number;
+}
+/** Audio and Azure Speech synchronization events emitted for one SSML request. */
+interface SsmlSynthesisResult {
+    audioData: ArrayBuffer;
+    durationMs: number;
+    boundaries?: SsmlSynthesisBoundary[];
+    /** Alias matching the Azure Speech event name. */
+    wordBoundary?: SsmlSynthesisBoundary[];
+    /** Alias for consumers that use Azure's word-boundary terminology. */
+    wordBoundaries?: SsmlSynthesisBoundary[];
+    visemes?: SsmlSynthesisViseme[];
+    bookmarks?: SsmlSynthesisBookmark[];
+}
 interface AzureTtsLogger {
     debug?: (...args: unknown[]) => void;
     info?: (...args: unknown[]) => void;
@@ -40,8 +65,11 @@ declare class AzureTtsClient {
     #private;
     constructor(options: AzureTtsClientOptions);
     synthesize(ssml: string): Promise<ArrayBuffer>;
+    synthesizeSsml(ssml: string): Promise<SsmlSynthesisResult>;
 }
 
+declare function synthesizeSsml(ssml: string, config: TtsConfig): Promise<SsmlSynthesisResult>;
+/** Backward-compatible audio-only synthesis helper. */
 declare function synthesizeSpeech(ssml: string, config: TtsConfig): Promise<ArrayBuffer>;
 
 interface FetchAzureVoiceCatalogOptions {
@@ -69,4 +97,4 @@ interface AzureVoiceCatalog {
 /** Fetches and deduplicates the current Azure Speech voice catalog for one or more regions. */
 declare function fetchAzureVoiceCatalog(options: FetchAzureVoiceCatalogOptions): Promise<AzureVoiceCatalog>;
 
-export { AzureTtsClient, type AzureTtsClientOptions, AzureTtsError, type AzureTtsLogger, AzureTtsSdkError, type AzureVoiceCatalog, type AzureVoiceCatalogVoice, type FetchAzureVoiceCatalogOptions, type FetchedAzureVoiceCatalogMetadata, type TtsConfig, fetchAzureVoiceCatalog, synthesizeSpeech };
+export { AzureTtsClient, type AzureTtsClientOptions, AzureTtsError, type AzureTtsLogger, AzureTtsSdkError, type AzureVoiceCatalog, type AzureVoiceCatalogVoice, type FetchAzureVoiceCatalogOptions, type FetchedAzureVoiceCatalogMetadata, type SsmlSynthesisBookmark, type SsmlSynthesisBoundary, type SsmlSynthesisResult, type SsmlSynthesisViseme, type TtsConfig, fetchAzureVoiceCatalog, synthesizeSpeech, synthesizeSsml };

package/dist/index.d.ts

@@ -1,4 +1,4 @@
-export { AudioElement, AzureDiagnosticCode, AzureLanguageNormalizationOptions, AzureSsmlValidationOptions, AzureValidationOptions, AzureVoiceCatalogMetadata, AzureVoiceDefinition, AzureVoiceMetadata, BookmarkElement, BreakElement, BuildPartialSsmlOptions, CustomElement, EmphasisElement, ExpressAsElement, ExtractSsmlTranslatableTextOptions, FromPlainTextToSsmlOptions, LangElement, LexiconElement, MapSsmlTextNodesOptions, MarkElement, MsttsAudioDurationElement, MsttsEmbeddingElement, MsttsSilenceElement, MsttsTtsEmbeddingElement, MsttsVisemeElement, MsttsVoiceConversionElement, NamedElement, ParagraphElement, PhonemeElement, ProsodyElement, SayAsElement, SentenceElement, SsmlAttributeValue, SsmlAttributes, SsmlBackgroundAudioNode, SsmlBreakElement, SsmlDiagnostic, SsmlDiagnosticSeverity, SsmlDiagnosticSource, SsmlDialogNode, SsmlDocument, SsmlElement, SsmlElementBase, SsmlEmbeddingNode, SsmlExpressAsElement, SsmlNode, SsmlPartialContext, SsmlPartialProsody, SsmlPartialVoice, SsmlPhonemeElement, SsmlProsodyElement, SsmlSayAsElement, SsmlStructureIntegrityResult, SsmlStructureMismatch, SsmlText, SsmlTextNodeContext, SsmlTtsEmbeddingNode, SsmlTurnNode, SsmlValidationError, SsmlVoiceConversionNode, SsmlVoiceElement, SubElement, VoiceElement, WordElement, areAzureLanguagesEquivalent, buildPartialSsml, buildSsml, extractSsmlText, extractSsmlTranslatableText, fromPlainTextToSsml, getAzureVoiceCatalogMetadata, getBuiltInVoiceCatalogMetadata, isValidAzureAudioDuration, mapSsmlTextNodes, normalizeAzureLanguage, parseSsml, validateAzureSsml, validateSsml, validateSsmlStructureIntegrity } from './core.js';
+export { AudioElement, AzureDiagnosticCode, AzureLanguageNormalizationOptions, AzureSsmlValidationOptions, AzureValidationOptions, AzureVoiceCatalogMetadata, AzureVoiceDefinition, AzureVoiceMetadata, BookmarkElement, BreakElement, BuildPartialSsmlOptions, CustomElement, EmphasisElement, ExpressAsElement, ExtractSsmlTranslatableTextOptions, FromPlainTextToSsmlOptions, LangElement, LexiconElement, MapSsmlTextNodesOptions, MarkElement, MsttsAudioDurationElement, MsttsEmbeddingElement, MsttsSilenceElement, MsttsTtsEmbeddingElement, MsttsVisemeElement, MsttsVoiceConversionElement, NamedElement, ParagraphElement, PhonemeElement, ProsodyElement, SayAsElement, SentenceElement, SsmlAttributeValue, SsmlAttributes, SsmlBackgroundAudioNode, SsmlBreakElement, SsmlDiagnostic, SsmlDiagnosticSeverity, SsmlDiagnosticSource, SsmlDialogNode, SsmlDocument, SsmlElement, SsmlElementBase, SsmlEmbeddingNode, SsmlExpressAsElement, SsmlNode, SsmlPartialContext, SsmlPartialProsody, SsmlPartialVoice, SsmlPhonemeElement, SsmlProsodyElement, SsmlSayAsElement, SsmlStructureIntegrityResult, SsmlStructureMismatch, SsmlText, SsmlTextNodeContext, SsmlTtsEmbeddingNode, SsmlTurnNode, SsmlValidationError, SsmlVoiceConversionNode, SsmlVoiceElement, SubElement, VoiceElement, WordElement, areAzureLanguagesEquivalent, buildPartialSsml, buildSsml, extractSsmlText, extractSsmlTranslatableText, fromPlainTextToSsml, getAzureVoiceCatalogMetadata, getBuiltInVoiceCatalogMetadata, isValidAzureAudioDuration, mapSsmlTextNodes, normalizeAzureLanguage, parseSsml, splitSsmlDocument, validateAzureSsml, validateSsml, validateSsmlStructureIntegrity } from './core.js';
 
 interface TtsConfig {
     signal?: AbortSignal;
@@ -8,6 +8,31 @@ interface TtsConfig {
     region: string;
     outputFormat?: string;
 }
+interface SsmlSynthesisBoundary {
+    text: string;
+    audioOffsetMs: number;
+    durationMs: number;
+}
+interface SsmlSynthesisViseme {
+    visemeId: number;
+    audioOffsetMs: number;
+}
+interface SsmlSynthesisBookmark {
+    name: string;
+    audioOffsetMs: number;
+}
+/** Audio and Azure Speech synchronization events emitted for one SSML request. */
+interface SsmlSynthesisResult {
+    audioData: ArrayBuffer;
+    durationMs: number;
+    boundaries?: SsmlSynthesisBoundary[];
+    /** Alias matching the Azure Speech event name. */
+    wordBoundary?: SsmlSynthesisBoundary[];
+    /** Alias for consumers that use Azure's word-boundary terminology. */
+    wordBoundaries?: SsmlSynthesisBoundary[];
+    visemes?: SsmlSynthesisViseme[];
+    bookmarks?: SsmlSynthesisBookmark[];
+}
 interface AzureTtsLogger {
     debug?: (...args: unknown[]) => void;
     info?: (...args: unknown[]) => void;
@@ -40,8 +65,11 @@ declare class AzureTtsClient {
     #private;
     constructor(options: AzureTtsClientOptions);
     synthesize(ssml: string): Promise<ArrayBuffer>;
+    synthesizeSsml(ssml: string): Promise<SsmlSynthesisResult>;
 }
 
+declare function synthesizeSsml(ssml: string, config: TtsConfig): Promise<SsmlSynthesisResult>;
+/** Backward-compatible audio-only synthesis helper. */
 declare function synthesizeSpeech(ssml: string, config: TtsConfig): Promise<ArrayBuffer>;
 
 interface FetchAzureVoiceCatalogOptions {
@@ -69,4 +97,4 @@ interface AzureVoiceCatalog {
 /** Fetches and deduplicates the current Azure Speech voice catalog for one or more regions. */
 declare function fetchAzureVoiceCatalog(options: FetchAzureVoiceCatalogOptions): Promise<AzureVoiceCatalog>;
 
-export { AzureTtsClient, type AzureTtsClientOptions, AzureTtsError, type AzureTtsLogger, AzureTtsSdkError, type AzureVoiceCatalog, type AzureVoiceCatalogVoice, type FetchAzureVoiceCatalogOptions, type FetchedAzureVoiceCatalogMetadata, type TtsConfig, fetchAzureVoiceCatalog, synthesizeSpeech };
+export { AzureTtsClient, type AzureTtsClientOptions, AzureTtsError, type AzureTtsLogger, AzureTtsSdkError, type AzureVoiceCatalog, type AzureVoiceCatalogVoice, type FetchAzureVoiceCatalogOptions, type FetchedAzureVoiceCatalogMetadata, type SsmlSynthesisBookmark, type SsmlSynthesisBoundary, type SsmlSynthesisResult, type SsmlSynthesisViseme, type TtsConfig, fetchAzureVoiceCatalog, synthesizeSpeech, synthesizeSsml };

package/dist/index.js

@@ -53,7 +53,9 @@ __export(src_exports, {
   mapSsmlTextNodes: () => mapSsmlTextNodes,
   normalizeAzureLanguage: () => normalizeAzureLanguage,
   parseSsml: () => parseSsml,
+  splitSsmlDocument: () => splitSsmlDocument,
   synthesizeSpeech: () => synthesizeSpeech,
+  synthesizeSsml: () => synthesizeSsml,
   validateAzureSsml: () => validateAzureSsml,
   validateSsml: () => validateSsml,
   validateSsmlStructureIntegrity: () => validateSsmlStructureIntegrity
@@ -987,6 +989,104 @@ function buildPartialSsml(textOrOptions, context) {
   return serializePartialSsml(textOrOptions.text, textOrOptions);
 }
 
+// packages/ssml-core/src/split.ts
+var DEFAULT_MAX_LENGTH = 1e4;
+function cloneElement(element, children) {
+  return { ...element, children };
+}
+function documentWithChildren(document, children) {
+  return buildSsml({ ...document, children });
+}
+function wrapWithContext(node, context) {
+  return context.reduceRight((current, parent) => cloneElement(parent, [current]), node);
+}
+function documentWithNode(document, node, context) {
+  return documentWithChildren(document, [wrapWithContext(node, context)]);
+}
+function splitTextNode(text, document, context, maxLength) {
+  if (!text) return [text];
+  const parts = [];
+  let start = 0;
+  while (start < text.length) {
+    let end = start + 1;
+    let bestEnd = end;
+    while (end <= text.length) {
+      if (documentWithNode(document, text.slice(start, end), context).length > maxLength) break;
+      bestEnd = end;
+      end += 1;
+    }
+    if (bestEnd === start) {
+      throw new RangeError("maxLength is too small to contain the SSML document wrapper");
+    }
+    const segment = text.slice(start, bestEnd);
+    const boundary = Math.max(segment.lastIndexOf(" "), segment.lastIndexOf("\n"), segment.lastIndexOf("	"));
+    const splitEnd = boundary > 0 ? start + boundary + 1 : bestEnd;
+    parts.push(text.slice(start, splitEnd));
+    start = splitEnd;
+  }
+  return parts;
+}
+function splitNode(document, node, maxLength, context = []) {
+  if (typeof node === "string" || node.type === "text") {
+    const value = typeof node === "string" ? node : node.value;
+    if (documentWithNode(document, node, context).length <= maxLength) return [node];
+    return splitTextNode(value, document, context, maxLength);
+  }
+  if (documentWithNode(document, node, context).length <= maxLength) return [node];
+  const children = node.children ?? [];
+  if (children.length === 0) {
+    throw new RangeError(`maxLength is too small to contain <${node.type}>`);
+  }
+  const splitChildren = children.flatMap((child) => splitNode(document, child, maxLength, [...context, node]));
+  const parts = [];
+  let group = [];
+  const flush = () => {
+    if (group.length > 0) {
+      parts.push(cloneElement(node, group));
+      group = [];
+    }
+  };
+  for (const child of splitChildren) {
+    const candidate = cloneElement(node, [...group, child]);
+    if (documentWithNode(document, candidate, context).length <= maxLength) {
+      group.push(child);
+      continue;
+    }
+    flush();
+    if (documentWithNode(document, cloneElement(node, [child]), context).length > maxLength) {
+      throw new RangeError(`maxLength is too small to contain <${node.type}>`);
+    }
+    group.push(child);
+  }
+  flush();
+  return parts;
+}
+function splitSsmlDocument(ssml, maxLength = DEFAULT_MAX_LENGTH) {
+  if (!Number.isInteger(maxLength) || maxLength <= 0) {
+    throw new RangeError("maxLength must be a positive integer");
+  }
+  const document = parseSsml(ssml);
+  if (ssml.length <= maxLength) return [ssml];
+  const children = document.children ?? [];
+  const splitChildren = children.flatMap((child) => splitNode(document, child, maxLength));
+  const chunks = [];
+  let group = [];
+  for (const child of splitChildren) {
+    const candidate = [...group, child];
+    if (documentWithChildren(document, candidate).length <= maxLength) {
+      group = candidate;
+      continue;
+    }
+    if (group.length > 0) chunks.push(group);
+    group = [child];
+    if (documentWithChildren(document, group).length > maxLength) {
+      throw new RangeError("maxLength is too small to contain the SSML document wrapper");
+    }
+  }
+  if (group.length > 0) chunks.push(group);
+  return chunks.map((chunk) => documentWithChildren(document, chunk));
+}
+
 // packages/ssml-core/src/validation.ts
 var PARSER_POSITION_SUFFIX = / at position (\d+)$/;
 function validateSsml(xmlString) {
@@ -1681,6 +1781,7 @@ function tokenizeElements(source) {
       attributes,
       childElementIndex,
       end,
+      depth: openElements.length + 1,
       name: tokenName,
       parentName: parent?.name,
       parentVoiceName,
@@ -1871,13 +1972,18 @@ function validateAudioSource(token, source, diagnostics, options, elementName2)
     addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> must be an absolute HTTP(S) URL.`);
     return;
   }
+  if (parsed.username || parsed.password)
+    addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> must not contain URL credentials.`);
   if (parsed.protocol !== "https:" && !(options.allowHttpAudio && parsed.protocol === "http:"))
     addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> must use HTTPS.`);
   const isAllowedOrigin = options.allowedAudioOrigins?.some((allowedOrigin) => {
     try {
-      return new URL(allowedOrigin).origin === parsed.origin;
+      const configured = new URL(allowedOrigin);
+      if (configured.protocol !== "https:" && configured.protocol !== "http:" || configured.username || configured.password || configured.pathname !== "/" || configured.search || configured.hash)
+        return false;
+      return configured.origin === parsed.origin;
     } catch {
-      return allowedOrigin === parsed.origin;
+      return false;
     }
   }) ?? false;
   if (options.allowedAudioOrigins && !isAllowedOrigin)
@@ -2085,6 +2191,9 @@ function validateAzureSsml(ssml, options = {}) {
   const maxLength = options.maxLength ?? 1e4;
   if (ssml.length > maxLength)
     addDiagnostic(diagnostics, ssml, maxLength, `SSML exceeds the maximum length of ${maxLength} characters.`);
+  if (options.maxXmlDepth !== void 0 && (!Number.isInteger(options.maxXmlDepth) || options.maxXmlDepth <= 0)) {
+    addDiagnostic(diagnostics, ssml, 0, "maxXmlDepth must be a positive integer.");
+  }
   try {
     parseSsml(ssml);
   } catch (error) {
@@ -2094,6 +2203,18 @@ function validateAzureSsml(ssml, options = {}) {
     return diagnostics;
   }
   const tokens = tokenizeElements(ssml);
+  if (options.maxXmlDepth !== void 0) {
+    for (const token of tokens) {
+      if (token.depth > options.maxXmlDepth) {
+        addDiagnostic(
+          diagnostics,
+          ssml,
+          token.start,
+          `XML nesting depth ${token.depth} exceeds the configured maximum of ${options.maxXmlDepth}.`
+        );
+      }
+    }
+  }
   const speak = tokens.find((token) => token.name.toLowerCase() === "speak");
   const voices = tokens.filter((token) => token.name.toLowerCase() === "voice");
   const backgroundAudioTokens = tokens.filter((token) => token.name.toLowerCase() === "mstts:backgroundaudio");
@@ -2302,7 +2423,8 @@ function closeSpeechResources(speechConfig, synthesizer) {
   } catch {
   }
 }
-async function synthesizeSpeech(ssml, config) {
+var ticksToMilliseconds = (ticks) => Math.max(0, ticks) / 1e4;
+async function synthesizeSsml(ssml, config) {
   if (config.signal?.aborted) {
     throw createSpeechSdkError("Speech synthesis was cancelled.");
   }
@@ -2329,6 +2451,22 @@ async function synthesizeSpeech(ssml, config) {
       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;
@@ -2340,7 +2478,20 @@ async function synthesizeSpeech(ssml, config) {
       settled = true;
       cleanup();
       closeResources();
-      resolve(result.audioData);
+      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;
+      resolve({
+        audioData: result.audioData,
+        durationMs,
+        ...boundaries.length > 0 ? { boundaries, wordBoundary: boundaries, wordBoundaries: boundaries } : {},
+        ...visemes.length > 0 ? { visemes } : {},
+        ...bookmarks.length > 0 ? { bookmarks } : {}
+      });
     };
     try {
       if (config.signal) {
@@ -2359,6 +2510,9 @@ async function synthesizeSpeech(ssml, config) {
     }
   });
 }
+async function synthesizeSpeech(ssml, config) {
+  return (await synthesizeSsml(ssml, config)).audioData;
+}
 
 // packages/azure-tts-client/src/client.ts
 var ENDPOINT_TEMPLATE = "https://{region}.tts.speech.microsoft.com/cognitiveservices/v1";
@@ -2375,6 +2529,12 @@ var AzureTtsClient = class {
     const config = { endpoint, region, subscriptionKey, outputFormat, signal, timeoutMs };
     return synthesizeSpeech(ssml, config);
   }
+  async synthesizeSsml(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);
+    return synthesizeSsml(ssml, { endpoint, region, subscriptionKey, outputFormat, signal, timeoutMs });
+  }
 };
 _options = new WeakMap();
 
@@ -2473,7 +2633,9 @@ async function fetchAzureVoiceCatalog(options) {
   mapSsmlTextNodes,
   normalizeAzureLanguage,
   parseSsml,
+  splitSsmlDocument,
   synthesizeSpeech,
+  synthesizeSsml,
   validateAzureSsml,
   validateSsml,
   validateSsmlStructureIntegrity