@ssml-builder-js/azure-tts-client 2.10.0 → 2.12.0

package/CHANGELOG.md

@@ -1,5 +1,17 @@
 # @ssml-builder-js/azure-tts-client
 
+## 2.12.0
+
+### Minor Changes
+
+- Add `synthesizeSsml` and `SsmlSynthesisResult` for word boundaries, visemes, bookmarks, and duration metadata.
+
+## 2.11.0
+
+### Minor Changes
+
+- Release v2.11.0 with Azure voice catalog APIs and synchronization tools, expanded editor voice support, and stricter background audio validation.
+
 ## 2.10.0
 
 ### Minor Changes

package/dist/index.d.mts

@@ -6,6 +6,31 @@ 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;
@@ -38,8 +63,36 @@ 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>;
 
-export { AzureTtsClient, type AzureTtsClientOptions, AzureTtsError, type AzureTtsLogger, AzureTtsSdkError, type TtsConfig, synthesizeSpeech };
+interface FetchAzureVoiceCatalogOptions {
+    apiKey: string;
+    region: string | string[];
+}
+interface AzureVoiceCatalogVoice {
+    name: string;
+    locale: string;
+    secondaryLocales?: readonly string[];
+    styles?: readonly string[];
+    regions: readonly string[];
+    status?: "ga" | "preview" | "deprecated";
+}
+interface FetchedAzureVoiceCatalogMetadata {
+    voiceCount: number;
+    generatedAt: string;
+    apiVersion: string;
+    regions: readonly string[];
+}
+interface AzureVoiceCatalog {
+    voices: readonly AzureVoiceCatalogVoice[];
+    metadata: FetchedAzureVoiceCatalogMetadata;
+}
+/** 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 SsmlSynthesisBookmark, type SsmlSynthesisBoundary, type SsmlSynthesisResult, type SsmlSynthesisViseme, type TtsConfig, fetchAzureVoiceCatalog, synthesizeSpeech, synthesizeSsml };

package/dist/index.d.ts

@@ -6,6 +6,31 @@ 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;
@@ -38,8 +63,36 @@ 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>;
 
-export { AzureTtsClient, type AzureTtsClientOptions, AzureTtsError, type AzureTtsLogger, AzureTtsSdkError, type TtsConfig, synthesizeSpeech };
+interface FetchAzureVoiceCatalogOptions {
+    apiKey: string;
+    region: string | string[];
+}
+interface AzureVoiceCatalogVoice {
+    name: string;
+    locale: string;
+    secondaryLocales?: readonly string[];
+    styles?: readonly string[];
+    regions: readonly string[];
+    status?: "ga" | "preview" | "deprecated";
+}
+interface FetchedAzureVoiceCatalogMetadata {
+    voiceCount: number;
+    generatedAt: string;
+    apiVersion: string;
+    regions: readonly string[];
+}
+interface AzureVoiceCatalog {
+    voices: readonly AzureVoiceCatalogVoice[];
+    metadata: FetchedAzureVoiceCatalogMetadata;
+}
+/** 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 SsmlSynthesisBookmark, type SsmlSynthesisBoundary, type SsmlSynthesisResult, type SsmlSynthesisViseme, type TtsConfig, fetchAzureVoiceCatalog, synthesizeSpeech, synthesizeSsml };

package/dist/index.js

@@ -40,7 +40,9 @@ __export(index_exports, {
   AzureTtsClient: () => AzureTtsClient,
   AzureTtsError: () => AzureTtsError,
   AzureTtsSdkError: () => AzureTtsSdkError,
-  synthesizeSpeech: () => synthesizeSpeech
+  fetchAzureVoiceCatalog: () => fetchAzureVoiceCatalog,
+  synthesizeSpeech: () => synthesizeSpeech,
+  synthesizeSsml: () => synthesizeSsml
 });
 module.exports = __toCommonJS(index_exports);
 
@@ -150,7 +152,8 @@ 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.");
   }
@@ -177,6 +180,22 @@ 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;
@@ -188,7 +207,20 @@ 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) {
@@ -207,6 +239,9 @@ async function synthesizeSpeech(ssml, config) {
     }
   });
 }
+async function synthesizeSpeech(ssml, config) {
+  return (await synthesizeSsml(ssml, config)).audioData;
+}
 
 // src/client.ts
 var ENDPOINT_TEMPLATE = "https://{region}.tts.speech.microsoft.com/cognitiveservices/v1";
@@ -223,13 +258,99 @@ 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();
+
+// src/voiceCatalog.ts
+var AZURE_VOICE_API_VERSION = "2025-10-01";
+function stringValue(value) {
+  return typeof value === "string" && value.trim() ? value.trim() : void 0;
+}
+function stringList(value) {
+  if (!Array.isArray(value)) return [];
+  return [...new Set(value.map(stringValue).filter((item) => item !== void 0))];
+}
+function normalizeStatus(value) {
+  const status = stringValue(value)?.toLowerCase();
+  if (status === "preview" || status === "deprecated" || status === "ga") return status;
+  return void 0;
+}
+function normalizeRegions(region) {
+  const regions = Array.isArray(region) ? region : [region];
+  const result = [...new Set(regions.map((item) => item.trim()).filter(Boolean))];
+  if (result.length === 0) throw new TypeError("At least one Azure Speech region is required.");
+  return result;
+}
+async function fetchRegionVoices(region, apiKey) {
+  const endpoint = `https://${encodeURIComponent(region)}.tts.speech.microsoft.com/cognitiveservices/voices/list`;
+  const response = await fetch(endpoint, {
+    headers: {
+      Accept: "application/json",
+      "Ocp-Apim-Subscription-Key": apiKey
+    }
+  });
+  if (!response.ok) {
+    throw new Error(`Azure List Voices API request failed for region "${region}" with HTTP ${response.status}.`);
+  }
+  const payload = await response.json();
+  if (!Array.isArray(payload)) throw new Error(`Azure List Voices API returned an invalid response for "${region}".`);
+  return payload.filter((item) => Boolean(item && typeof item === "object"));
+}
+async function fetchAzureVoiceCatalog(options) {
+  if (!options || typeof options.apiKey !== "string" || !options.apiKey.trim())
+    throw new TypeError("An Azure Speech API key is required.");
+  const regions = normalizeRegions(options.region);
+  const payloads = await Promise.all(regions.map((region) => fetchRegionVoices(region, options.apiKey)));
+  const voices = /* @__PURE__ */ new Map();
+  for (let regionIndex = 0; regionIndex < payloads.length; regionIndex += 1) {
+    const region = regions[regionIndex];
+    for (const record of payloads[regionIndex]) {
+      const name = stringValue(record.ShortName) ?? stringValue(record.Name);
+      const locale = stringValue(record.Locale);
+      if (!name || !locale) continue;
+      const key = name.toLowerCase();
+      const existing = voices.get(key);
+      const secondaryLocales = stringList(record.SecondaryLocaleList);
+      const styles = stringList(record.StyleList);
+      const status = normalizeStatus(record.Status);
+      const merged = {
+        name: existing?.name ?? name,
+        locale: existing?.locale ?? locale,
+        regions: [.../* @__PURE__ */ new Set([...existing?.regions ?? [], region])]
+      };
+      const mergedSecondaryLocales = [.../* @__PURE__ */ new Set([...existing?.secondaryLocales ?? [], ...secondaryLocales])];
+      if (mergedSecondaryLocales.length > 0) merged.secondaryLocales = mergedSecondaryLocales;
+      const mergedStyles = [.../* @__PURE__ */ new Set([...existing?.styles ?? [], ...styles])];
+      if (mergedStyles.length > 0) merged.styles = mergedStyles;
+      if (status) merged.status = status;
+      else if (existing?.status) merged.status = existing.status;
+      voices.set(key, merged);
+    }
+  }
+  const sortedVoices = [...voices.values()].sort((first, second) => first.name.localeCompare(second.name));
+  return {
+    voices: sortedVoices,
+    metadata: {
+      voiceCount: sortedVoices.length,
+      generatedAt: (/* @__PURE__ */ new Date()).toISOString(),
+      apiVersion: AZURE_VOICE_API_VERSION,
+      regions
+    }
+  };
+}
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {
   AzureTtsClient,
   AzureTtsError,
   AzureTtsSdkError,
-  synthesizeSpeech
+  fetchAzureVoiceCatalog,
+  synthesizeSpeech,
+  synthesizeSsml
 });
 //# sourceMappingURL=index.js.map

package/dist/index.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/index.ts","../src/errors.ts","../src/synthesis.ts","../src/speechConfig.ts","../src/outputFormats.ts","../src/client.ts"],"sourcesContent":["/**\n * azure-tts-client: Azure Text-to-Speech client for SSML playback.\n */\n\nexport type { AzureTtsClientOptions, AzureTtsLogger, TtsConfig } from \"./types.ts\";\nexport { AzureTtsError, AzureTtsSdkError } from \"./errors.ts\";\nexport { AzureTtsClient } from \"./client.ts\";\nexport { synthesizeSpeech } from \"./synthesis.ts\";\n","export class AzureTtsError extends Error {\n  readonly status: number;\n  readonly statusText: string;\n  readonly responseBody: string;\n  readonly requestId: string | null;\n\n  constructor(status: number, statusText: string, responseBody: string, requestId: string | null) {\n    super(`Azure TTS request failed: ${status} ${statusText}`);\n    this.name = \"AzureTtsError\";\n    this.status = status;\n    this.statusText = statusText;\n    this.responseBody = responseBody;\n    this.requestId = requestId;\n  }\n}\n\nexport class AzureTtsSdkError extends AzureTtsError {\n  readonly errorDetails: string;\n\n  constructor(errorDetails: string) {\n    super(0, \"Speech SDK\", errorDetails, null);\n    this.name = \"AzureTtsSdkError\";\n    this.message = `Azure TTS synthesis failed: ${errorDetails}`;\n    this.errorDetails = errorDetails;\n  }\n}\n\nexport function createSpeechSdkError(error: unknown): AzureTtsSdkError {\n  const message = error instanceof Error ? error.message : String(error);\n  return new AzureTtsSdkError(message);\n}\n","import * as SpeechSDK from \"microsoft-cognitiveservices-speech-sdk\";\nimport { createSpeechSdkError } from \"./errors.ts\";\nimport { createSpeechConfig } from \"./speechConfig.ts\";\nimport type { TtsConfig } from \"./types.ts\";\n\nfunction closeSpeechResources(speechConfig: SpeechSDK.SpeechConfig, synthesizer: SpeechSDK.SpeechSynthesizer): void {\n  try {\n    synthesizer.close();\n  } catch {}\n\n  try {\n    speechConfig.close();\n  } catch {}\n}\n\nexport async function synthesizeSpeech(ssml: string, config: TtsConfig): Promise<ArrayBuffer> {\n  if (config.signal?.aborted) {\n    throw createSpeechSdkError(\"Speech synthesis was cancelled.\");\n  }\n\n  const speechConfig = createSpeechConfig(config);\n  const synthesizer = new SpeechSDK.SpeechSynthesizer(speechConfig, null);\n\n  return await new Promise<ArrayBuffer>((resolve, reject) => {\n    let resourcesClosed = false;\n    let settled = false;\n    let timeout: ReturnType<typeof setTimeout> | undefined;\n    let abortHandler: (() => void) | undefined;\n    const cleanup = () => {\n      if (timeout) clearTimeout(timeout);\n      if (abortHandler) config.signal?.removeEventListener(\"abort\", abortHandler);\n    };\n    const closeResources = () => {\n      if (resourcesClosed) return;\n      resourcesClosed = true;\n      closeSpeechResources(speechConfig, synthesizer);\n    };\n    const rejectWithError = (error: unknown) => {\n      if (settled) return;\n      settled = true;\n      cleanup();\n      closeResources();\n      reject(createSpeechSdkError(error));\n    };\n\n    const cb = (result: SpeechSDK.SpeechSynthesisResult) => {\n      if (settled) return;\n      const { reason, errorDetails } = result;\n      if (reason !== SpeechSDK.ResultReason.SynthesizingAudioCompleted) {\n        const err = errorDetails || `Speech synthesis failed with reason ${reason}.`;\n        rejectWithError(err);\n        return;\n      }\n      settled = true;\n      cleanup();\n      closeResources();\n      resolve(result.audioData);\n    };\n\n    try {\n      if (config.signal) {\n        abortHandler = () => rejectWithError(\"Speech synthesis was cancelled.\");\n        config.signal.addEventListener(\"abort\", abortHandler, { once: true });\n      }\n      if (config.timeoutMs !== undefined && config.timeoutMs > 0) {\n        timeout = setTimeout(\n          () => rejectWithError(`Speech synthesis timed out after ${config.timeoutMs} ms.`),\n          config.timeoutMs,\n        );\n      }\n      synthesizer.speakSsmlAsync(ssml, cb, rejectWithError);\n    } catch (error) {\n      rejectWithError(error);\n    }\n  });\n}\n","import { SpeechConfig } from \"microsoft-cognitiveservices-speech-sdk\";\nimport { DEFAULT_OUTPUT_FORMAT, resolveOutputFormat } from \"./outputFormats.ts\";\nimport type { TtsConfig } from \"./types.ts\";\n\nexport function resolveEndpoint(config: TtsConfig): string {\n  const endpoint = config.endpoint?.trim() || \"https://{region}.tts.speech.microsoft.com/cognitiveservices/v1\";\n  return endpoint.replace(/\\{region\\}/g, encodeURIComponent(config.region));\n}\n\nexport function createSpeechConfig(config: TtsConfig): SpeechConfig {\n  const { outputFormat = DEFAULT_OUTPUT_FORMAT, subscriptionKey } = config;\n\n  const endpoint = new URL(resolveEndpoint(config));\n  const speechConfig = SpeechConfig.fromEndpoint(endpoint, subscriptionKey);\n  speechConfig.speechSynthesisOutputFormat = resolveOutputFormat(outputFormat);\n  return speechConfig;\n}\n","import * as SpeechSDK from \"microsoft-cognitiveservices-speech-sdk\";\n\nexport const DEFAULT_OUTPUT_FORMAT = \"audio-16khz-128kbitrate-mono-mp3\";\n\nconst OUTPUT_FORMATS: Record<string, SpeechSDK.SpeechSynthesisOutputFormat> = {\n  \"raw-8khz-8bit-mono-mulaw\": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz8BitMonoMULaw,\n  \"riff-16khz-16kbps-mono-siren\": SpeechSDK.SpeechSynthesisOutputFormat.Riff16Khz16KbpsMonoSiren,\n  \"audio-16khz-16kbps-mono-siren\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz16KbpsMonoSiren,\n  \"audio-16khz-32kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz32KBitRateMonoMp3,\n  \"audio-16khz-128kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz128KBitRateMonoMp3,\n  \"audio-16khz-64kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz64KBitRateMonoMp3,\n  \"audio-24khz-48kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz48KBitRateMonoMp3,\n  \"audio-24khz-96kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz96KBitRateMonoMp3,\n  \"audio-24khz-160kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz160KBitRateMonoMp3,\n  \"raw-16khz-16bit-mono-truesilk\": SpeechSDK.SpeechSynthesisOutputFormat.Raw16Khz16BitMonoTrueSilk,\n  \"riff-16khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff16Khz16BitMonoPcm,\n  \"riff-8khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz16BitMonoPcm,\n  \"riff-24khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff24Khz16BitMonoPcm,\n  \"riff-8khz-8bit-mono-mulaw\": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz8BitMonoMULaw,\n  \"raw-16khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw16Khz16BitMonoPcm,\n  \"raw-24khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw24Khz16BitMonoPcm,\n  \"raw-8khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz16BitMonoPcm,\n  \"ogg-16khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Ogg16Khz16BitMonoOpus,\n  \"ogg-24khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Ogg24Khz16BitMonoOpus,\n  \"raw-48khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw48Khz16BitMonoPcm,\n  \"riff-48khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff48Khz16BitMonoPcm,\n  \"audio-48khz-96kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio48Khz96KBitRateMonoMp3,\n  \"audio-48khz-192kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio48Khz192KBitRateMonoMp3,\n  \"ogg-48khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Ogg48Khz16BitMonoOpus,\n  \"webm-16khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Webm16Khz16BitMonoOpus,\n  \"webm-24khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Webm24Khz16BitMonoOpus,\n  \"webm-24khz-16bit-24kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Webm24Khz16Bit24KbpsMonoOpus,\n  \"raw-24khz-16bit-mono-truesilk\": SpeechSDK.SpeechSynthesisOutputFormat.Raw24Khz16BitMonoTrueSilk,\n  \"raw-8khz-8bit-mono-alaw\": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz8BitMonoALaw,\n  \"riff-8khz-8bit-mono-alaw\": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz8BitMonoALaw,\n  \"audio-16khz-16bit-32kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz16Bit32KbpsMonoOpus,\n  \"audio-24khz-16bit-48kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz16Bit48KbpsMonoOpus,\n  \"audio-24khz-16bit-24kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz16Bit24KbpsMonoOpus,\n  \"raw-22050hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw22050Hz16BitMonoPcm,\n  \"riff-22050hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff22050Hz16BitMonoPcm,\n  \"raw-44100hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw44100Hz16BitMonoPcm,\n  \"riff-44100hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff44100Hz16BitMonoPcm,\n  \"amr-wb-16000hz\": SpeechSDK.SpeechSynthesisOutputFormat.AmrWb16000Hz,\n  \"g722-16khz-64kbps\": SpeechSDK.SpeechSynthesisOutputFormat.G72216Khz64Kbps,\n};\n\nexport function resolveOutputFormat(outputFormat: string): SpeechSDK.SpeechSynthesisOutputFormat {\n  const resolvedFormat = OUTPUT_FORMATS[outputFormat];\n  if (resolvedFormat === undefined) {\n    throw new Error(`Unsupported Azure Speech output format: ${outputFormat}`);\n  }\n\n  return resolvedFormat;\n}\n","import { synthesizeSpeech } from \"./synthesis.ts\";\nimport type { AzureTtsClientOptions } from \"./types.ts\";\n\nconst ENDPOINT_TEMPLATE = \"https://{region}.tts.speech.microsoft.com/cognitiveservices/v1\";\n\nexport class AzureTtsClient {\n  readonly #options: AzureTtsClientOptions;\n\n  constructor(options: AzureTtsClientOptions) {\n    this.#options = options;\n  }\n\n  async synthesize(ssml: string): Promise<ArrayBuffer> {\n    const { region, subscriptionKey, outputFormat, signal, timeoutMs } = this.#options;\n    const endpoint = this.#options.endpoint?.trim() || ENDPOINT_TEMPLATE.replace(\"{region}\", region);\n    this.#options.logger?.debug?.(\"Using Azure TTS endpoint:\", endpoint);\n\n    const config = { endpoint, region, subscriptionKey, outputFormat, signal, timeoutMs };\n    return synthesizeSpeech(ssml, config);\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;;;ACAO,IAAM,gBAAN,cAA4B,MAAM;AAAA,EAMvC,YAAY,QAAgB,YAAoB,cAAsB,WAA0B;AAC9F,UAAM,6BAA6B,MAAM,IAAI,UAAU,EAAE;AACzD,SAAK,OAAO;AACZ,SAAK,SAAS;AACd,SAAK,aAAa;AAClB,SAAK,eAAe;AACpB,SAAK,YAAY;AAAA,EACnB;AACF;AAEO,IAAM,mBAAN,cAA+B,cAAc;AAAA,EAGlD,YAAY,cAAsB;AAChC,UAAM,GAAG,cAAc,cAAc,IAAI;AACzC,SAAK,OAAO;AACZ,SAAK,UAAU,+BAA+B,YAAY;AAC1D,SAAK,eAAe;AAAA,EACtB;AACF;AAEO,SAAS,qBAAqB,OAAkC;AACrE,QAAM,UAAU,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AACrE,SAAO,IAAI,iBAAiB,OAAO;AACrC;;;AC9BA,IAAAA,aAA2B;;;ACA3B,oDAA6B;;;ACA7B,gBAA2B;AAEpB,IAAM,wBAAwB;AAErC,IAAM,iBAAwE;AAAA,EAC5E,4BAAsC,sCAA4B;AAAA,EAClE,gCAA0C,sCAA4B;AAAA,EACtE,iCAA2C,sCAA4B;AAAA,EACvE,mCAA6C,sCAA4B;AAAA,EACzE,oCAA8C,sCAA4B;AAAA,EAC1E,mCAA6C,sCAA4B;AAAA,EACzE,mCAA6C,sCAA4B;AAAA,EACzE,mCAA6C,sCAA4B;AAAA,EACzE,oCAA8C,sCAA4B;AAAA,EAC1E,iCAA2C,sCAA4B;AAAA,EACvE,6BAAuC,sCAA4B;AAAA,EACnE,4BAAsC,sCAA4B;AAAA,EAClE,6BAAuC,sCAA4B;AAAA,EACnE,6BAAuC,sCAA4B;AAAA,EACnE,4BAAsC,sCAA4B;AAAA,EAClE,4BAAsC,sCAA4B;AAAA,EAClE,2BAAqC,sCAA4B;AAAA,EACjE,6BAAuC,sCAA4B;AAAA,EACnE,6BAAuC,sCAA4B;AAAA,EACnE,4BAAsC,sCAA4B;AAAA,EAClE,6BAAuC,sCAA4B;AAAA,EACnE,mCAA6C,sCAA4B;AAAA,EACzE,oCAA8C,sCAA4B;AAAA,EAC1E,6BAAuC,sCAA4B;AAAA,EACnE,8BAAwC,sCAA4B;AAAA,EACpE,8BAAwC,sCAA4B;AAAA,EACpE,qCAA+C,sCAA4B;AAAA,EAC3E,iCAA2C,sCAA4B;AAAA,EACvE,2BAAqC,sCAA4B;AAAA,EACjE,4BAAsC,sCAA4B;AAAA,EAClE,sCAAgD,sCAA4B;AAAA,EAC5E,sCAAgD,sCAA4B;AAAA,EAC5E,sCAAgD,sCAA4B;AAAA,EAC5E,8BAAwC,sCAA4B;AAAA,EACpE,+BAAyC,sCAA4B;AAAA,EACrE,8BAAwC,sCAA4B;AAAA,EACpE,+BAAyC,sCAA4B;AAAA,EACrE,kBAA4B,sCAA4B;AAAA,EACxD,qBAA+B,sCAA4B;AAC7D;AAEO,SAAS,oBAAoB,cAA6D;AAC/F,QAAM,iBAAiB,eAAe,YAAY;AAClD,MAAI,mBAAmB,QAAW;AAChC,UAAM,IAAI,MAAM,2CAA2C,YAAY,EAAE;AAAA,EAC3E;AAEA,SAAO;AACT;;;ADjDO,SAAS,gBAAgB,QAA2B;AACzD,QAAM,WAAW,OAAO,UAAU,KAAK,KAAK;AAC5C,SAAO,SAAS,QAAQ,eAAe,mBAAmB,OAAO,MAAM,CAAC;AAC1E;AAEO,SAAS,mBAAmB,QAAiC;AAClE,QAAM,EAAE,eAAe,uBAAuB,gBAAgB,IAAI;AAElE,QAAM,WAAW,IAAI,IAAI,gBAAgB,MAAM,CAAC;AAChD,QAAM,eAAe,2DAAa,aAAa,UAAU,eAAe;AACxE,eAAa,8BAA8B,oBAAoB,YAAY;AAC3E,SAAO;AACT;;;ADXA,SAAS,qBAAqB,cAAsC,aAAgD;AAClH,MAAI;AACF,gBAAY,MAAM;AAAA,EACpB,QAAQ;AAAA,EAAC;AAET,MAAI;AACF,iBAAa,MAAM;AAAA,EACrB,QAAQ;AAAA,EAAC;AACX;AAEA,eAAsB,iBAAiB,MAAc,QAAyC;AAC5F,MAAI,OAAO,QAAQ,SAAS;AAC1B,UAAM,qBAAqB,iCAAiC;AAAA,EAC9D;AAEA,QAAM,eAAe,mBAAmB,MAAM;AAC9C,QAAM,cAAc,IAAc,6BAAkB,cAAc,IAAI;AAEtE,SAAO,MAAM,IAAI,QAAqB,CAAC,SAAS,WAAW;AACzD,QAAI,kBAAkB;AACtB,QAAI,UAAU;AACd,QAAI;AACJ,QAAI;AACJ,UAAM,UAAU,MAAM;AACpB,UAAI,QAAS,cAAa,OAAO;AACjC,UAAI,aAAc,QAAO,QAAQ,oBAAoB,SAAS,YAAY;AAAA,IAC5E;AACA,UAAM,iBAAiB,MAAM;AAC3B,UAAI,gBAAiB;AACrB,wBAAkB;AAClB,2BAAqB,cAAc,WAAW;AAAA,IAChD;AACA,UAAM,kBAAkB,CAAC,UAAmB;AAC1C,UAAI,QAAS;AACb,gBAAU;AACV,cAAQ;AACR,qBAAe;AACf,aAAO,qBAAqB,KAAK,CAAC;AAAA,IACpC;AAEA,UAAM,KAAK,CAAC,WAA4C;AACtD,UAAI,QAAS;AACb,YAAM,EAAE,QAAQ,aAAa,IAAI;AACjC,UAAI,WAAqB,wBAAa,4BAA4B;AAChE,cAAM,MAAM,gBAAgB,uCAAuC,MAAM;AACzE,wBAAgB,GAAG;AACnB;AAAA,MACF;AACA,gBAAU;AACV,cAAQ;AACR,qBAAe;AACf,cAAQ,OAAO,SAAS;AAAA,IAC1B;AAEA,QAAI;AACF,UAAI,OAAO,QAAQ;AACjB,uBAAe,MAAM,gBAAgB,iCAAiC;AACtE,eAAO,OAAO,iBAAiB,SAAS,cAAc,EAAE,MAAM,KAAK,CAAC;AAAA,MACtE;AACA,UAAI,OAAO,cAAc,UAAa,OAAO,YAAY,GAAG;AAC1D,kBAAU;AAAA,UACR,MAAM,gBAAgB,oCAAoC,OAAO,SAAS,MAAM;AAAA,UAChF,OAAO;AAAA,QACT;AAAA,MACF;AACA,kBAAY,eAAe,MAAM,IAAI,eAAe;AAAA,IACtD,SAAS,OAAO;AACd,sBAAgB,KAAK;AAAA,IACvB;AAAA,EACF,CAAC;AACH;;;AGxEA,IAAM,oBAAoB;AAH1B;AAKO,IAAM,iBAAN,MAAqB;AAAA,EAG1B,YAAY,SAAgC;AAF5C,uBAAS;AAGP,uBAAK,UAAW;AAAA,EAClB;AAAA,EAEA,MAAM,WAAW,MAAoC;AACnD,UAAM,EAAE,QAAQ,iBAAiB,cAAc,QAAQ,UAAU,IAAI,mBAAK;AAC1E,UAAM,WAAW,mBAAK,UAAS,UAAU,KAAK,KAAK,kBAAkB,QAAQ,YAAY,MAAM;AAC/F,uBAAK,UAAS,QAAQ,QAAQ,6BAA6B,QAAQ;AAEnE,UAAM,SAAS,EAAE,UAAU,QAAQ,iBAAiB,cAAc,QAAQ,UAAU;AACpF,WAAO,iBAAiB,MAAM,MAAM;AAAA,EACtC;AACF;AAdW;","names":["SpeechSDK"]}
+{"version":3,"sources":["../src/index.ts","../src/errors.ts","../src/synthesis.ts","../src/speechConfig.ts","../src/outputFormats.ts","../src/client.ts","../src/voiceCatalog.ts"],"sourcesContent":["/**\n * azure-tts-client: Azure Text-to-Speech client for SSML playback.\n */\n\nexport type {\n  AzureTtsClientOptions,\n  AzureTtsLogger,\n  SsmlSynthesisBookmark,\n  SsmlSynthesisBoundary,\n  SsmlSynthesisResult,\n  SsmlSynthesisViseme,\n  TtsConfig,\n} from \"./types.ts\";\nexport { AzureTtsError, AzureTtsSdkError } from \"./errors.ts\";\nexport { AzureTtsClient } from \"./client.ts\";\nexport { synthesizeSpeech } from \"./synthesis.ts\";\nexport { synthesizeSsml } from \"./synthesis.ts\";\nexport { fetchAzureVoiceCatalog } from \"./voiceCatalog.ts\";\nexport type {\n  AzureVoiceCatalog,\n  AzureVoiceCatalogVoice,\n  FetchedAzureVoiceCatalogMetadata,\n  FetchAzureVoiceCatalogOptions,\n} from \"./voiceCatalog.ts\";\n","export class AzureTtsError extends Error {\n  readonly status: number;\n  readonly statusText: string;\n  readonly responseBody: string;\n  readonly requestId: string | null;\n\n  constructor(status: number, statusText: string, responseBody: string, requestId: string | null) {\n    super(`Azure TTS request failed: ${status} ${statusText}`);\n    this.name = \"AzureTtsError\";\n    this.status = status;\n    this.statusText = statusText;\n    this.responseBody = responseBody;\n    this.requestId = requestId;\n  }\n}\n\nexport class AzureTtsSdkError extends AzureTtsError {\n  readonly errorDetails: string;\n\n  constructor(errorDetails: string) {\n    super(0, \"Speech SDK\", errorDetails, null);\n    this.name = \"AzureTtsSdkError\";\n    this.message = `Azure TTS synthesis failed: ${errorDetails}`;\n    this.errorDetails = errorDetails;\n  }\n}\n\nexport function createSpeechSdkError(error: unknown): AzureTtsSdkError {\n  const message = error instanceof Error ? error.message : String(error);\n  return new AzureTtsSdkError(message);\n}\n","import * as SpeechSDK from \"microsoft-cognitiveservices-speech-sdk\";\nimport { createSpeechSdkError } from \"./errors.ts\";\nimport { createSpeechConfig } from \"./speechConfig.ts\";\nimport type { SsmlSynthesisResult, TtsConfig } from \"./types.ts\";\n\nfunction closeSpeechResources(speechConfig: SpeechSDK.SpeechConfig, synthesizer: SpeechSDK.SpeechSynthesizer): void {\n  try {\n    synthesizer.close();\n  } catch {}\n\n  try {\n    speechConfig.close();\n  } catch {}\n}\n\nconst ticksToMilliseconds = (ticks: number): number => Math.max(0, ticks) / 10_000;\n\nexport async function synthesizeSsml(ssml: string, config: TtsConfig): Promise<SsmlSynthesisResult> {\n  if (config.signal?.aborted) {\n    throw createSpeechSdkError(\"Speech synthesis was cancelled.\");\n  }\n\n  const speechConfig = createSpeechConfig(config);\n  const synthesizer = new SpeechSDK.SpeechSynthesizer(speechConfig, null);\n\n  return await new Promise<SsmlSynthesisResult>((resolve, reject) => {\n    let resourcesClosed = false;\n    let settled = false;\n    let timeout: ReturnType<typeof setTimeout> | undefined;\n    let abortHandler: (() => void) | undefined;\n    const cleanup = () => {\n      if (timeout) clearTimeout(timeout);\n      if (abortHandler) config.signal?.removeEventListener(\"abort\", abortHandler);\n    };\n    const closeResources = () => {\n      if (resourcesClosed) return;\n      resourcesClosed = true;\n      closeSpeechResources(speechConfig, synthesizer);\n    };\n    const rejectWithError = (error: unknown) => {\n      if (settled) return;\n      settled = true;\n      cleanup();\n      closeResources();\n      reject(createSpeechSdkError(error));\n    };\n\n    const boundaries: SsmlSynthesisResult[\"boundaries\"] = [];\n    const visemes: SsmlSynthesisResult[\"visemes\"] = [];\n    const bookmarks: SsmlSynthesisResult[\"bookmarks\"] = [];\n    synthesizer.wordBoundary = (_sender, event) => {\n      boundaries.push({\n        text: event.text,\n        audioOffsetMs: ticksToMilliseconds(event.audioOffset),\n        durationMs: ticksToMilliseconds(event.duration),\n      });\n    };\n    synthesizer.visemeReceived = (_sender, event) => {\n      visemes.push({ visemeId: event.visemeId, audioOffsetMs: ticksToMilliseconds(event.audioOffset) });\n    };\n    synthesizer.bookmarkReached = (_sender, event) => {\n      bookmarks.push({ name: event.text, audioOffsetMs: ticksToMilliseconds(event.audioOffset) });\n    };\n\n    const cb = (result: SpeechSDK.SpeechSynthesisResult) => {\n      if (settled) return;\n      const { reason, errorDetails } = result;\n      if (reason !== SpeechSDK.ResultReason.SynthesizingAudioCompleted) {\n        const err = errorDetails || `Speech synthesis failed with reason ${reason}.`;\n        rejectWithError(err);\n        return;\n      }\n      settled = true;\n      cleanup();\n      closeResources();\n      const eventDurationMs = Math.max(\n        0,\n        ...(boundaries ?? []).map((boundary) => boundary.audioOffsetMs + boundary.durationMs),\n        ...(visemes ?? []).map((viseme) => viseme.audioOffsetMs),\n        ...(bookmarks ?? []).map((bookmark) => bookmark.audioOffsetMs),\n      );\n      const durationMs = result.audioDuration ? ticksToMilliseconds(result.audioDuration) : eventDurationMs;\n      resolve({\n        audioData: result.audioData,\n        durationMs,\n        ...(boundaries.length > 0 ? { boundaries, wordBoundary: boundaries, wordBoundaries: boundaries } : {}),\n        ...(visemes.length > 0 ? { visemes } : {}),\n        ...(bookmarks.length > 0 ? { bookmarks } : {}),\n      });\n    };\n\n    try {\n      if (config.signal) {\n        abortHandler = () => rejectWithError(\"Speech synthesis was cancelled.\");\n        config.signal.addEventListener(\"abort\", abortHandler, { once: true });\n      }\n      if (config.timeoutMs !== undefined && config.timeoutMs > 0) {\n        timeout = setTimeout(\n          () => rejectWithError(`Speech synthesis timed out after ${config.timeoutMs} ms.`),\n          config.timeoutMs,\n        );\n      }\n      synthesizer.speakSsmlAsync(ssml, cb, rejectWithError);\n    } catch (error) {\n      rejectWithError(error);\n    }\n  });\n}\n\n/** Backward-compatible audio-only synthesis helper. */\nexport async function synthesizeSpeech(ssml: string, config: TtsConfig): Promise<ArrayBuffer> {\n  return (await synthesizeSsml(ssml, config)).audioData;\n}\n","import { SpeechConfig } from \"microsoft-cognitiveservices-speech-sdk\";\nimport { DEFAULT_OUTPUT_FORMAT, resolveOutputFormat } from \"./outputFormats.ts\";\nimport type { TtsConfig } from \"./types.ts\";\n\nexport function resolveEndpoint(config: TtsConfig): string {\n  const endpoint = config.endpoint?.trim() || \"https://{region}.tts.speech.microsoft.com/cognitiveservices/v1\";\n  return endpoint.replace(/\\{region\\}/g, encodeURIComponent(config.region));\n}\n\nexport function createSpeechConfig(config: TtsConfig): SpeechConfig {\n  const { outputFormat = DEFAULT_OUTPUT_FORMAT, subscriptionKey } = config;\n\n  const endpoint = new URL(resolveEndpoint(config));\n  const speechConfig = SpeechConfig.fromEndpoint(endpoint, subscriptionKey);\n  speechConfig.speechSynthesisOutputFormat = resolveOutputFormat(outputFormat);\n  return speechConfig;\n}\n","import * as SpeechSDK from \"microsoft-cognitiveservices-speech-sdk\";\n\nexport const DEFAULT_OUTPUT_FORMAT = \"audio-16khz-128kbitrate-mono-mp3\";\n\nconst OUTPUT_FORMATS: Record<string, SpeechSDK.SpeechSynthesisOutputFormat> = {\n  \"raw-8khz-8bit-mono-mulaw\": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz8BitMonoMULaw,\n  \"riff-16khz-16kbps-mono-siren\": SpeechSDK.SpeechSynthesisOutputFormat.Riff16Khz16KbpsMonoSiren,\n  \"audio-16khz-16kbps-mono-siren\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz16KbpsMonoSiren,\n  \"audio-16khz-32kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz32KBitRateMonoMp3,\n  \"audio-16khz-128kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz128KBitRateMonoMp3,\n  \"audio-16khz-64kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz64KBitRateMonoMp3,\n  \"audio-24khz-48kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz48KBitRateMonoMp3,\n  \"audio-24khz-96kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz96KBitRateMonoMp3,\n  \"audio-24khz-160kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz160KBitRateMonoMp3,\n  \"raw-16khz-16bit-mono-truesilk\": SpeechSDK.SpeechSynthesisOutputFormat.Raw16Khz16BitMonoTrueSilk,\n  \"riff-16khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff16Khz16BitMonoPcm,\n  \"riff-8khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz16BitMonoPcm,\n  \"riff-24khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff24Khz16BitMonoPcm,\n  \"riff-8khz-8bit-mono-mulaw\": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz8BitMonoMULaw,\n  \"raw-16khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw16Khz16BitMonoPcm,\n  \"raw-24khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw24Khz16BitMonoPcm,\n  \"raw-8khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz16BitMonoPcm,\n  \"ogg-16khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Ogg16Khz16BitMonoOpus,\n  \"ogg-24khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Ogg24Khz16BitMonoOpus,\n  \"raw-48khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw48Khz16BitMonoPcm,\n  \"riff-48khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff48Khz16BitMonoPcm,\n  \"audio-48khz-96kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio48Khz96KBitRateMonoMp3,\n  \"audio-48khz-192kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio48Khz192KBitRateMonoMp3,\n  \"ogg-48khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Ogg48Khz16BitMonoOpus,\n  \"webm-16khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Webm16Khz16BitMonoOpus,\n  \"webm-24khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Webm24Khz16BitMonoOpus,\n  \"webm-24khz-16bit-24kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Webm24Khz16Bit24KbpsMonoOpus,\n  \"raw-24khz-16bit-mono-truesilk\": SpeechSDK.SpeechSynthesisOutputFormat.Raw24Khz16BitMonoTrueSilk,\n  \"raw-8khz-8bit-mono-alaw\": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz8BitMonoALaw,\n  \"riff-8khz-8bit-mono-alaw\": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz8BitMonoALaw,\n  \"audio-16khz-16bit-32kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz16Bit32KbpsMonoOpus,\n  \"audio-24khz-16bit-48kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz16Bit48KbpsMonoOpus,\n  \"audio-24khz-16bit-24kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz16Bit24KbpsMonoOpus,\n  \"raw-22050hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw22050Hz16BitMonoPcm,\n  \"riff-22050hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff22050Hz16BitMonoPcm,\n  \"raw-44100hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw44100Hz16BitMonoPcm,\n  \"riff-44100hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff44100Hz16BitMonoPcm,\n  \"amr-wb-16000hz\": SpeechSDK.SpeechSynthesisOutputFormat.AmrWb16000Hz,\n  \"g722-16khz-64kbps\": SpeechSDK.SpeechSynthesisOutputFormat.G72216Khz64Kbps,\n};\n\nexport function resolveOutputFormat(outputFormat: string): SpeechSDK.SpeechSynthesisOutputFormat {\n  const resolvedFormat = OUTPUT_FORMATS[outputFormat];\n  if (resolvedFormat === undefined) {\n    throw new Error(`Unsupported Azure Speech output format: ${outputFormat}`);\n  }\n\n  return resolvedFormat;\n}\n","import { synthesizeSpeech, synthesizeSsml } from \"./synthesis.ts\";\nimport type { AzureTtsClientOptions, SsmlSynthesisResult } from \"./types.ts\";\n\nconst ENDPOINT_TEMPLATE = \"https://{region}.tts.speech.microsoft.com/cognitiveservices/v1\";\n\nexport class AzureTtsClient {\n  readonly #options: AzureTtsClientOptions;\n\n  constructor(options: AzureTtsClientOptions) {\n    this.#options = options;\n  }\n\n  async synthesize(ssml: string): Promise<ArrayBuffer> {\n    const { region, subscriptionKey, outputFormat, signal, timeoutMs } = this.#options;\n    const endpoint = this.#options.endpoint?.trim() || ENDPOINT_TEMPLATE.replace(\"{region}\", region);\n    this.#options.logger?.debug?.(\"Using Azure TTS endpoint:\", endpoint);\n\n    const config = { endpoint, region, subscriptionKey, outputFormat, signal, timeoutMs };\n    return synthesizeSpeech(ssml, config);\n  }\n\n  async synthesizeSsml(ssml: string): Promise<SsmlSynthesisResult> {\n    const { region, subscriptionKey, outputFormat, signal, timeoutMs } = this.#options;\n    const endpoint = this.#options.endpoint?.trim() || ENDPOINT_TEMPLATE.replace(\"{region}\", region);\n    this.#options.logger?.debug?.(\"Using Azure TTS endpoint:\", endpoint);\n\n    return synthesizeSsml(ssml, { endpoint, region, subscriptionKey, outputFormat, signal, timeoutMs });\n  }\n}\n","const AZURE_VOICE_API_VERSION = \"2025-10-01\";\n\nexport interface FetchAzureVoiceCatalogOptions {\n  apiKey: string;\n  region: string | string[];\n}\n\nexport interface AzureVoiceCatalogVoice {\n  name: string;\n  locale: string;\n  secondaryLocales?: readonly string[];\n  styles?: readonly string[];\n  regions: readonly string[];\n  status?: \"ga\" | \"preview\" | \"deprecated\";\n}\n\nexport interface FetchedAzureVoiceCatalogMetadata {\n  voiceCount: number;\n  generatedAt: string;\n  apiVersion: string;\n  regions: readonly string[];\n}\n\nexport interface AzureVoiceCatalog {\n  voices: readonly AzureVoiceCatalogVoice[];\n  metadata: FetchedAzureVoiceCatalogMetadata;\n}\n\ninterface AzureVoiceApiRecord {\n  Locale?: unknown;\n  Name?: unknown;\n  SecondaryLocaleList?: unknown;\n  ShortName?: unknown;\n  Status?: unknown;\n  StyleList?: unknown;\n}\n\nfunction stringValue(value: unknown): string | undefined {\n  return typeof value === \"string\" && value.trim() ? value.trim() : undefined;\n}\n\nfunction stringList(value: unknown): string[] {\n  if (!Array.isArray(value)) return [];\n  return [...new Set(value.map(stringValue).filter((item): item is string => item !== undefined))];\n}\n\nfunction normalizeStatus(value: unknown): AzureVoiceCatalogVoice[\"status\"] {\n  const status = stringValue(value)?.toLowerCase();\n  if (status === \"preview\" || status === \"deprecated\" || status === \"ga\") return status;\n  return undefined;\n}\n\nfunction normalizeRegions(region: string | string[]): string[] {\n  const regions = Array.isArray(region) ? region : [region];\n  const result = [...new Set(regions.map((item) => item.trim()).filter(Boolean))];\n  if (result.length === 0) throw new TypeError(\"At least one Azure Speech region is required.\");\n  return result;\n}\n\nasync function fetchRegionVoices(region: string, apiKey: string): Promise<AzureVoiceApiRecord[]> {\n  const endpoint = `https://${encodeURIComponent(region)}.tts.speech.microsoft.com/cognitiveservices/voices/list`;\n  const response = await fetch(endpoint, {\n    headers: {\n      Accept: \"application/json\",\n      \"Ocp-Apim-Subscription-Key\": apiKey,\n    },\n  });\n  if (!response.ok) {\n    throw new Error(`Azure List Voices API request failed for region \"${region}\" with HTTP ${response.status}.`);\n  }\n  const payload: unknown = await response.json();\n  if (!Array.isArray(payload)) throw new Error(`Azure List Voices API returned an invalid response for \"${region}\".`);\n  return payload.filter((item): item is AzureVoiceApiRecord => Boolean(item && typeof item === \"object\"));\n}\n\n/** Fetches and deduplicates the current Azure Speech voice catalog for one or more regions. */\nexport async function fetchAzureVoiceCatalog(options: FetchAzureVoiceCatalogOptions): Promise<AzureVoiceCatalog> {\n  if (!options || typeof options.apiKey !== \"string\" || !options.apiKey.trim())\n    throw new TypeError(\"An Azure Speech API key is required.\");\n  const regions = normalizeRegions(options.region);\n  const payloads = await Promise.all(regions.map((region) => fetchRegionVoices(region, options.apiKey)));\n  const voices = new Map<string, AzureVoiceCatalogVoice>();\n\n  for (let regionIndex = 0; regionIndex < payloads.length; regionIndex += 1) {\n    const region = regions[regionIndex];\n    for (const record of payloads[regionIndex]) {\n      const name = stringValue(record.ShortName) ?? stringValue(record.Name);\n      const locale = stringValue(record.Locale);\n      if (!name || !locale) continue;\n      const key = name.toLowerCase();\n      const existing = voices.get(key);\n      const secondaryLocales = stringList(record.SecondaryLocaleList);\n      const styles = stringList(record.StyleList);\n      const status = normalizeStatus(record.Status);\n      const merged: AzureVoiceCatalogVoice = {\n        name: existing?.name ?? name,\n        locale: existing?.locale ?? locale,\n        regions: [...new Set([...(existing?.regions ?? []), region])],\n      };\n      const mergedSecondaryLocales = [...new Set([...(existing?.secondaryLocales ?? []), ...secondaryLocales])];\n      if (mergedSecondaryLocales.length > 0) merged.secondaryLocales = mergedSecondaryLocales;\n      const mergedStyles = [...new Set([...(existing?.styles ?? []), ...styles])];\n      if (mergedStyles.length > 0) merged.styles = mergedStyles;\n      if (status) merged.status = status;\n      else if (existing?.status) merged.status = existing.status;\n      voices.set(key, merged);\n    }\n  }\n\n  const sortedVoices = [...voices.values()].sort((first, second) => first.name.localeCompare(second.name));\n  return {\n    voices: sortedVoices,\n    metadata: {\n      voiceCount: sortedVoices.length,\n      generatedAt: new Date().toISOString(),\n      apiVersion: AZURE_VOICE_API_VERSION,\n      regions,\n    },\n  };\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;;;ACAO,IAAM,gBAAN,cAA4B,MAAM;AAAA,EAMvC,YAAY,QAAgB,YAAoB,cAAsB,WAA0B;AAC9F,UAAM,6BAA6B,MAAM,IAAI,UAAU,EAAE;AACzD,SAAK,OAAO;AACZ,SAAK,SAAS;AACd,SAAK,aAAa;AAClB,SAAK,eAAe;AACpB,SAAK,YAAY;AAAA,EACnB;AACF;AAEO,IAAM,mBAAN,cAA+B,cAAc;AAAA,EAGlD,YAAY,cAAsB;AAChC,UAAM,GAAG,cAAc,cAAc,IAAI;AACzC,SAAK,OAAO;AACZ,SAAK,UAAU,+BAA+B,YAAY;AAC1D,SAAK,eAAe;AAAA,EACtB;AACF;AAEO,SAAS,qBAAqB,OAAkC;AACrE,QAAM,UAAU,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AACrE,SAAO,IAAI,iBAAiB,OAAO;AACrC;;;AC9BA,IAAAA,aAA2B;;;ACA3B,oDAA6B;;;ACA7B,gBAA2B;AAEpB,IAAM,wBAAwB;AAErC,IAAM,iBAAwE;AAAA,EAC5E,4BAAsC,sCAA4B;AAAA,EAClE,gCAA0C,sCAA4B;AAAA,EACtE,iCAA2C,sCAA4B;AAAA,EACvE,mCAA6C,sCAA4B;AAAA,EACzE,oCAA8C,sCAA4B;AAAA,EAC1E,mCAA6C,sCAA4B;AAAA,EACzE,mCAA6C,sCAA4B;AAAA,EACzE,mCAA6C,sCAA4B;AAAA,EACzE,oCAA8C,sCAA4B;AAAA,EAC1E,iCAA2C,sCAA4B;AAAA,EACvE,6BAAuC,sCAA4B;AAAA,EACnE,4BAAsC,sCAA4B;AAAA,EAClE,6BAAuC,sCAA4B;AAAA,EACnE,6BAAuC,sCAA4B;AAAA,EACnE,4BAAsC,sCAA4B;AAAA,EAClE,4BAAsC,sCAA4B;AAAA,EAClE,2BAAqC,sCAA4B;AAAA,EACjE,6BAAuC,sCAA4B;AAAA,EACnE,6BAAuC,sCAA4B;AAAA,EACnE,4BAAsC,sCAA4B;AAAA,EAClE,6BAAuC,sCAA4B;AAAA,EACnE,mCAA6C,sCAA4B;AAAA,EACzE,oCAA8C,sCAA4B;AAAA,EAC1E,6BAAuC,sCAA4B;AAAA,EACnE,8BAAwC,sCAA4B;AAAA,EACpE,8BAAwC,sCAA4B;AAAA,EACpE,qCAA+C,sCAA4B;AAAA,EAC3E,iCAA2C,sCAA4B;AAAA,EACvE,2BAAqC,sCAA4B;AAAA,EACjE,4BAAsC,sCAA4B;AAAA,EAClE,sCAAgD,sCAA4B;AAAA,EAC5E,sCAAgD,sCAA4B;AAAA,EAC5E,sCAAgD,sCAA4B;AAAA,EAC5E,8BAAwC,sCAA4B;AAAA,EACpE,+BAAyC,sCAA4B;AAAA,EACrE,8BAAwC,sCAA4B;AAAA,EACpE,+BAAyC,sCAA4B;AAAA,EACrE,kBAA4B,sCAA4B;AAAA,EACxD,qBAA+B,sCAA4B;AAC7D;AAEO,SAAS,oBAAoB,cAA6D;AAC/F,QAAM,iBAAiB,eAAe,YAAY;AAClD,MAAI,mBAAmB,QAAW;AAChC,UAAM,IAAI,MAAM,2CAA2C,YAAY,EAAE;AAAA,EAC3E;AAEA,SAAO;AACT;;;ADjDO,SAAS,gBAAgB,QAA2B;AACzD,QAAM,WAAW,OAAO,UAAU,KAAK,KAAK;AAC5C,SAAO,SAAS,QAAQ,eAAe,mBAAmB,OAAO,MAAM,CAAC;AAC1E;AAEO,SAAS,mBAAmB,QAAiC;AAClE,QAAM,EAAE,eAAe,uBAAuB,gBAAgB,IAAI;AAElE,QAAM,WAAW,IAAI,IAAI,gBAAgB,MAAM,CAAC;AAChD,QAAM,eAAe,2DAAa,aAAa,UAAU,eAAe;AACxE,eAAa,8BAA8B,oBAAoB,YAAY;AAC3E,SAAO;AACT;;;ADXA,SAAS,qBAAqB,cAAsC,aAAgD;AAClH,MAAI;AACF,gBAAY,MAAM;AAAA,EACpB,QAAQ;AAAA,EAAC;AAET,MAAI;AACF,iBAAa,MAAM;AAAA,EACrB,QAAQ;AAAA,EAAC;AACX;AAEA,IAAM,sBAAsB,CAAC,UAA0B,KAAK,IAAI,GAAG,KAAK,IAAI;AAE5E,eAAsB,eAAe,MAAc,QAAiD;AAClG,MAAI,OAAO,QAAQ,SAAS;AAC1B,UAAM,qBAAqB,iCAAiC;AAAA,EAC9D;AAEA,QAAM,eAAe,mBAAmB,MAAM;AAC9C,QAAM,cAAc,IAAc,6BAAkB,cAAc,IAAI;AAEtE,SAAO,MAAM,IAAI,QAA6B,CAAC,SAAS,WAAW;AACjE,QAAI,kBAAkB;AACtB,QAAI,UAAU;AACd,QAAI;AACJ,QAAI;AACJ,UAAM,UAAU,MAAM;AACpB,UAAI,QAAS,cAAa,OAAO;AACjC,UAAI,aAAc,QAAO,QAAQ,oBAAoB,SAAS,YAAY;AAAA,IAC5E;AACA,UAAM,iBAAiB,MAAM;AAC3B,UAAI,gBAAiB;AACrB,wBAAkB;AAClB,2BAAqB,cAAc,WAAW;AAAA,IAChD;AACA,UAAM,kBAAkB,CAAC,UAAmB;AAC1C,UAAI,QAAS;AACb,gBAAU;AACV,cAAQ;AACR,qBAAe;AACf,aAAO,qBAAqB,KAAK,CAAC;AAAA,IACpC;AAEA,UAAM,aAAgD,CAAC;AACvD,UAAM,UAA0C,CAAC;AACjD,UAAM,YAA8C,CAAC;AACrD,gBAAY,eAAe,CAAC,SAAS,UAAU;AAC7C,iBAAW,KAAK;AAAA,QACd,MAAM,MAAM;AAAA,QACZ,eAAe,oBAAoB,MAAM,WAAW;AAAA,QACpD,YAAY,oBAAoB,MAAM,QAAQ;AAAA,MAChD,CAAC;AAAA,IACH;AACA,gBAAY,iBAAiB,CAAC,SAAS,UAAU;AAC/C,cAAQ,KAAK,EAAE,UAAU,MAAM,UAAU,eAAe,oBAAoB,MAAM,WAAW,EAAE,CAAC;AAAA,IAClG;AACA,gBAAY,kBAAkB,CAAC,SAAS,UAAU;AAChD,gBAAU,KAAK,EAAE,MAAM,MAAM,MAAM,eAAe,oBAAoB,MAAM,WAAW,EAAE,CAAC;AAAA,IAC5F;AAEA,UAAM,KAAK,CAAC,WAA4C;AACtD,UAAI,QAAS;AACb,YAAM,EAAE,QAAQ,aAAa,IAAI;AACjC,UAAI,WAAqB,wBAAa,4BAA4B;AAChE,cAAM,MAAM,gBAAgB,uCAAuC,MAAM;AACzE,wBAAgB,GAAG;AACnB;AAAA,MACF;AACA,gBAAU;AACV,cAAQ;AACR,qBAAe;AACf,YAAM,kBAAkB,KAAK;AAAA,QAC3B;AAAA,QACA,IAAI,cAAc,CAAC,GAAG,IAAI,CAAC,aAAa,SAAS,gBAAgB,SAAS,UAAU;AAAA,QACpF,IAAI,WAAW,CAAC,GAAG,IAAI,CAAC,WAAW,OAAO,aAAa;AAAA,QACvD,IAAI,aAAa,CAAC,GAAG,IAAI,CAAC,aAAa,SAAS,aAAa;AAAA,MAC/D;AACA,YAAM,aAAa,OAAO,gBAAgB,oBAAoB,OAAO,aAAa,IAAI;AACtF,cAAQ;AAAA,QACN,WAAW,OAAO;AAAA,QAClB;AAAA,QACA,GAAI,WAAW,SAAS,IAAI,EAAE,YAAY,cAAc,YAAY,gBAAgB,WAAW,IAAI,CAAC;AAAA,QACpG,GAAI,QAAQ,SAAS,IAAI,EAAE,QAAQ,IAAI,CAAC;AAAA,QACxC,GAAI,UAAU,SAAS,IAAI,EAAE,UAAU,IAAI,CAAC;AAAA,MAC9C,CAAC;AAAA,IACH;AAEA,QAAI;AACF,UAAI,OAAO,QAAQ;AACjB,uBAAe,MAAM,gBAAgB,iCAAiC;AACtE,eAAO,OAAO,iBAAiB,SAAS,cAAc,EAAE,MAAM,KAAK,CAAC;AAAA,MACtE;AACA,UAAI,OAAO,cAAc,UAAa,OAAO,YAAY,GAAG;AAC1D,kBAAU;AAAA,UACR,MAAM,gBAAgB,oCAAoC,OAAO,SAAS,MAAM;AAAA,UAChF,OAAO;AAAA,QACT;AAAA,MACF;AACA,kBAAY,eAAe,MAAM,IAAI,eAAe;AAAA,IACtD,SAAS,OAAO;AACd,sBAAgB,KAAK;AAAA,IACvB;AAAA,EACF,CAAC;AACH;AAGA,eAAsB,iBAAiB,MAAc,QAAyC;AAC5F,UAAQ,MAAM,eAAe,MAAM,MAAM,GAAG;AAC9C;;;AG7GA,IAAM,oBAAoB;AAH1B;AAKO,IAAM,iBAAN,MAAqB;AAAA,EAG1B,YAAY,SAAgC;AAF5C,uBAAS;AAGP,uBAAK,UAAW;AAAA,EAClB;AAAA,EAEA,MAAM,WAAW,MAAoC;AACnD,UAAM,EAAE,QAAQ,iBAAiB,cAAc,QAAQ,UAAU,IAAI,mBAAK;AAC1E,UAAM,WAAW,mBAAK,UAAS,UAAU,KAAK,KAAK,kBAAkB,QAAQ,YAAY,MAAM;AAC/F,uBAAK,UAAS,QAAQ,QAAQ,6BAA6B,QAAQ;AAEnE,UAAM,SAAS,EAAE,UAAU,QAAQ,iBAAiB,cAAc,QAAQ,UAAU;AACpF,WAAO,iBAAiB,MAAM,MAAM;AAAA,EACtC;AAAA,EAEA,MAAM,eAAe,MAA4C;AAC/D,UAAM,EAAE,QAAQ,iBAAiB,cAAc,QAAQ,UAAU,IAAI,mBAAK;AAC1E,UAAM,WAAW,mBAAK,UAAS,UAAU,KAAK,KAAK,kBAAkB,QAAQ,YAAY,MAAM;AAC/F,uBAAK,UAAS,QAAQ,QAAQ,6BAA6B,QAAQ;AAEnE,WAAO,eAAe,MAAM,EAAE,UAAU,QAAQ,iBAAiB,cAAc,QAAQ,UAAU,CAAC;AAAA,EACpG;AACF;AAtBW;;;ACNX,IAAM,0BAA0B;AAqChC,SAAS,YAAY,OAAoC;AACvD,SAAO,OAAO,UAAU,YAAY,MAAM,KAAK,IAAI,MAAM,KAAK,IAAI;AACpE;AAEA,SAAS,WAAW,OAA0B;AAC5C,MAAI,CAAC,MAAM,QAAQ,KAAK,EAAG,QAAO,CAAC;AACnC,SAAO,CAAC,GAAG,IAAI,IAAI,MAAM,IAAI,WAAW,EAAE,OAAO,CAAC,SAAyB,SAAS,MAAS,CAAC,CAAC;AACjG;AAEA,SAAS,gBAAgB,OAAkD;AACzE,QAAM,SAAS,YAAY,KAAK,GAAG,YAAY;AAC/C,MAAI,WAAW,aAAa,WAAW,gBAAgB,WAAW,KAAM,QAAO;AAC/E,SAAO;AACT;AAEA,SAAS,iBAAiB,QAAqC;AAC7D,QAAM,UAAU,MAAM,QAAQ,MAAM,IAAI,SAAS,CAAC,MAAM;AACxD,QAAM,SAAS,CAAC,GAAG,IAAI,IAAI,QAAQ,IAAI,CAAC,SAAS,KAAK,KAAK,CAAC,EAAE,OAAO,OAAO,CAAC,CAAC;AAC9E,MAAI,OAAO,WAAW,EAAG,OAAM,IAAI,UAAU,+CAA+C;AAC5F,SAAO;AACT;AAEA,eAAe,kBAAkB,QAAgB,QAAgD;AAC/F,QAAM,WAAW,WAAW,mBAAmB,MAAM,CAAC;AACtD,QAAM,WAAW,MAAM,MAAM,UAAU;AAAA,IACrC,SAAS;AAAA,MACP,QAAQ;AAAA,MACR,6BAA6B;AAAA,IAC/B;AAAA,EACF,CAAC;AACD,MAAI,CAAC,SAAS,IAAI;AAChB,UAAM,IAAI,MAAM,oDAAoD,MAAM,eAAe,SAAS,MAAM,GAAG;AAAA,EAC7G;AACA,QAAM,UAAmB,MAAM,SAAS,KAAK;AAC7C,MAAI,CAAC,MAAM,QAAQ,OAAO,EAAG,OAAM,IAAI,MAAM,2DAA2D,MAAM,IAAI;AAClH,SAAO,QAAQ,OAAO,CAAC,SAAsC,QAAQ,QAAQ,OAAO,SAAS,QAAQ,CAAC;AACxG;AAGA,eAAsB,uBAAuB,SAAoE;AAC/G,MAAI,CAAC,WAAW,OAAO,QAAQ,WAAW,YAAY,CAAC,QAAQ,OAAO,KAAK;AACzE,UAAM,IAAI,UAAU,sCAAsC;AAC5D,QAAM,UAAU,iBAAiB,QAAQ,MAAM;AAC/C,QAAM,WAAW,MAAM,QAAQ,IAAI,QAAQ,IAAI,CAAC,WAAW,kBAAkB,QAAQ,QAAQ,MAAM,CAAC,CAAC;AACrG,QAAM,SAAS,oBAAI,IAAoC;AAEvD,WAAS,cAAc,GAAG,cAAc,SAAS,QAAQ,eAAe,GAAG;AACzE,UAAM,SAAS,QAAQ,WAAW;AAClC,eAAW,UAAU,SAAS,WAAW,GAAG;AAC1C,YAAM,OAAO,YAAY,OAAO,SAAS,KAAK,YAAY,OAAO,IAAI;AACrE,YAAM,SAAS,YAAY,OAAO,MAAM;AACxC,UAAI,CAAC,QAAQ,CAAC,OAAQ;AACtB,YAAM,MAAM,KAAK,YAAY;AAC7B,YAAM,WAAW,OAAO,IAAI,GAAG;AAC/B,YAAM,mBAAmB,WAAW,OAAO,mBAAmB;AAC9D,YAAM,SAAS,WAAW,OAAO,SAAS;AAC1C,YAAM,SAAS,gBAAgB,OAAO,MAAM;AAC5C,YAAM,SAAiC;AAAA,QACrC,MAAM,UAAU,QAAQ;AAAA,QACxB,QAAQ,UAAU,UAAU;AAAA,QAC5B,SAAS,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAI,UAAU,WAAW,CAAC,GAAI,MAAM,CAAC,CAAC;AAAA,MAC9D;AACA,YAAM,yBAAyB,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAI,UAAU,oBAAoB,CAAC,GAAI,GAAG,gBAAgB,CAAC,CAAC;AACxG,UAAI,uBAAuB,SAAS,EAAG,QAAO,mBAAmB;AACjE,YAAM,eAAe,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAI,UAAU,UAAU,CAAC,GAAI,GAAG,MAAM,CAAC,CAAC;AAC1E,UAAI,aAAa,SAAS,EAAG,QAAO,SAAS;AAC7C,UAAI,OAAQ,QAAO,SAAS;AAAA,eACnB,UAAU,OAAQ,QAAO,SAAS,SAAS;AACpD,aAAO,IAAI,KAAK,MAAM;AAAA,IACxB;AAAA,EACF;AAEA,QAAM,eAAe,CAAC,GAAG,OAAO,OAAO,CAAC,EAAE,KAAK,CAAC,OAAO,WAAW,MAAM,KAAK,cAAc,OAAO,IAAI,CAAC;AACvG,SAAO;AAAA,IACL,QAAQ;AAAA,IACR,UAAU;AAAA,MACR,YAAY,aAAa;AAAA,MACzB,cAAa,oBAAI,KAAK,GAAE,YAAY;AAAA,MACpC,YAAY;AAAA,MACZ;AAAA,IACF;AAAA,EACF;AACF;","names":["SpeechSDK"]}

package/dist/index.mjs

@@ -112,7 +112,8 @@ 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.");
   }
@@ -139,6 +140,22 @@ 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;
@@ -150,7 +167,20 @@ 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) {
@@ -169,6 +199,9 @@ async function synthesizeSpeech(ssml, config) {
     }
   });
 }
+async function synthesizeSpeech(ssml, config) {
+  return (await synthesizeSsml(ssml, config)).audioData;
+}
 
 // src/client.ts
 var ENDPOINT_TEMPLATE = "https://{region}.tts.speech.microsoft.com/cognitiveservices/v1";
@@ -185,12 +218,98 @@ 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();
+
+// src/voiceCatalog.ts
+var AZURE_VOICE_API_VERSION = "2025-10-01";
+function stringValue(value) {
+  return typeof value === "string" && value.trim() ? value.trim() : void 0;
+}
+function stringList(value) {
+  if (!Array.isArray(value)) return [];
+  return [...new Set(value.map(stringValue).filter((item) => item !== void 0))];
+}
+function normalizeStatus(value) {
+  const status = stringValue(value)?.toLowerCase();
+  if (status === "preview" || status === "deprecated" || status === "ga") return status;
+  return void 0;
+}
+function normalizeRegions(region) {
+  const regions = Array.isArray(region) ? region : [region];
+  const result = [...new Set(regions.map((item) => item.trim()).filter(Boolean))];
+  if (result.length === 0) throw new TypeError("At least one Azure Speech region is required.");
+  return result;
+}
+async function fetchRegionVoices(region, apiKey) {
+  const endpoint = `https://${encodeURIComponent(region)}.tts.speech.microsoft.com/cognitiveservices/voices/list`;
+  const response = await fetch(endpoint, {
+    headers: {
+      Accept: "application/json",
+      "Ocp-Apim-Subscription-Key": apiKey
+    }
+  });
+  if (!response.ok) {
+    throw new Error(`Azure List Voices API request failed for region "${region}" with HTTP ${response.status}.`);
+  }
+  const payload = await response.json();
+  if (!Array.isArray(payload)) throw new Error(`Azure List Voices API returned an invalid response for "${region}".`);
+  return payload.filter((item) => Boolean(item && typeof item === "object"));
+}
+async function fetchAzureVoiceCatalog(options) {
+  if (!options || typeof options.apiKey !== "string" || !options.apiKey.trim())
+    throw new TypeError("An Azure Speech API key is required.");
+  const regions = normalizeRegions(options.region);
+  const payloads = await Promise.all(regions.map((region) => fetchRegionVoices(region, options.apiKey)));
+  const voices = /* @__PURE__ */ new Map();
+  for (let regionIndex = 0; regionIndex < payloads.length; regionIndex += 1) {
+    const region = regions[regionIndex];
+    for (const record of payloads[regionIndex]) {
+      const name = stringValue(record.ShortName) ?? stringValue(record.Name);
+      const locale = stringValue(record.Locale);
+      if (!name || !locale) continue;
+      const key = name.toLowerCase();
+      const existing = voices.get(key);
+      const secondaryLocales = stringList(record.SecondaryLocaleList);
+      const styles = stringList(record.StyleList);
+      const status = normalizeStatus(record.Status);
+      const merged = {
+        name: existing?.name ?? name,
+        locale: existing?.locale ?? locale,
+        regions: [.../* @__PURE__ */ new Set([...existing?.regions ?? [], region])]
+      };
+      const mergedSecondaryLocales = [.../* @__PURE__ */ new Set([...existing?.secondaryLocales ?? [], ...secondaryLocales])];
+      if (mergedSecondaryLocales.length > 0) merged.secondaryLocales = mergedSecondaryLocales;
+      const mergedStyles = [.../* @__PURE__ */ new Set([...existing?.styles ?? [], ...styles])];
+      if (mergedStyles.length > 0) merged.styles = mergedStyles;
+      if (status) merged.status = status;
+      else if (existing?.status) merged.status = existing.status;
+      voices.set(key, merged);
+    }
+  }
+  const sortedVoices = [...voices.values()].sort((first, second) => first.name.localeCompare(second.name));
+  return {
+    voices: sortedVoices,
+    metadata: {
+      voiceCount: sortedVoices.length,
+      generatedAt: (/* @__PURE__ */ new Date()).toISOString(),
+      apiVersion: AZURE_VOICE_API_VERSION,
+      regions
+    }
+  };
+}
 export {
   AzureTtsClient,
   AzureTtsError,
   AzureTtsSdkError,
-  synthesizeSpeech
+  fetchAzureVoiceCatalog,
+  synthesizeSpeech,
+  synthesizeSsml
 };
 //# sourceMappingURL=index.mjs.map

package/dist/index.mjs.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/errors.ts","../src/synthesis.ts","../src/speechConfig.ts","../src/outputFormats.ts","../src/client.ts"],"sourcesContent":["export class AzureTtsError extends Error {\n  readonly status: number;\n  readonly statusText: string;\n  readonly responseBody: string;\n  readonly requestId: string | null;\n\n  constructor(status: number, statusText: string, responseBody: string, requestId: string | null) {\n    super(`Azure TTS request failed: ${status} ${statusText}`);\n    this.name = \"AzureTtsError\";\n    this.status = status;\n    this.statusText = statusText;\n    this.responseBody = responseBody;\n    this.requestId = requestId;\n  }\n}\n\nexport class AzureTtsSdkError extends AzureTtsError {\n  readonly errorDetails: string;\n\n  constructor(errorDetails: string) {\n    super(0, \"Speech SDK\", errorDetails, null);\n    this.name = \"AzureTtsSdkError\";\n    this.message = `Azure TTS synthesis failed: ${errorDetails}`;\n    this.errorDetails = errorDetails;\n  }\n}\n\nexport function createSpeechSdkError(error: unknown): AzureTtsSdkError {\n  const message = error instanceof Error ? error.message : String(error);\n  return new AzureTtsSdkError(message);\n}\n","import * as SpeechSDK from \"microsoft-cognitiveservices-speech-sdk\";\nimport { createSpeechSdkError } from \"./errors.ts\";\nimport { createSpeechConfig } from \"./speechConfig.ts\";\nimport type { TtsConfig } from \"./types.ts\";\n\nfunction closeSpeechResources(speechConfig: SpeechSDK.SpeechConfig, synthesizer: SpeechSDK.SpeechSynthesizer): void {\n  try {\n    synthesizer.close();\n  } catch {}\n\n  try {\n    speechConfig.close();\n  } catch {}\n}\n\nexport async function synthesizeSpeech(ssml: string, config: TtsConfig): Promise<ArrayBuffer> {\n  if (config.signal?.aborted) {\n    throw createSpeechSdkError(\"Speech synthesis was cancelled.\");\n  }\n\n  const speechConfig = createSpeechConfig(config);\n  const synthesizer = new SpeechSDK.SpeechSynthesizer(speechConfig, null);\n\n  return await new Promise<ArrayBuffer>((resolve, reject) => {\n    let resourcesClosed = false;\n    let settled = false;\n    let timeout: ReturnType<typeof setTimeout> | undefined;\n    let abortHandler: (() => void) | undefined;\n    const cleanup = () => {\n      if (timeout) clearTimeout(timeout);\n      if (abortHandler) config.signal?.removeEventListener(\"abort\", abortHandler);\n    };\n    const closeResources = () => {\n      if (resourcesClosed) return;\n      resourcesClosed = true;\n      closeSpeechResources(speechConfig, synthesizer);\n    };\n    const rejectWithError = (error: unknown) => {\n      if (settled) return;\n      settled = true;\n      cleanup();\n      closeResources();\n      reject(createSpeechSdkError(error));\n    };\n\n    const cb = (result: SpeechSDK.SpeechSynthesisResult) => {\n      if (settled) return;\n      const { reason, errorDetails } = result;\n      if (reason !== SpeechSDK.ResultReason.SynthesizingAudioCompleted) {\n        const err = errorDetails || `Speech synthesis failed with reason ${reason}.`;\n        rejectWithError(err);\n        return;\n      }\n      settled = true;\n      cleanup();\n      closeResources();\n      resolve(result.audioData);\n    };\n\n    try {\n      if (config.signal) {\n        abortHandler = () => rejectWithError(\"Speech synthesis was cancelled.\");\n        config.signal.addEventListener(\"abort\", abortHandler, { once: true });\n      }\n      if (config.timeoutMs !== undefined && config.timeoutMs > 0) {\n        timeout = setTimeout(\n          () => rejectWithError(`Speech synthesis timed out after ${config.timeoutMs} ms.`),\n          config.timeoutMs,\n        );\n      }\n      synthesizer.speakSsmlAsync(ssml, cb, rejectWithError);\n    } catch (error) {\n      rejectWithError(error);\n    }\n  });\n}\n","import { SpeechConfig } from \"microsoft-cognitiveservices-speech-sdk\";\nimport { DEFAULT_OUTPUT_FORMAT, resolveOutputFormat } from \"./outputFormats.ts\";\nimport type { TtsConfig } from \"./types.ts\";\n\nexport function resolveEndpoint(config: TtsConfig): string {\n  const endpoint = config.endpoint?.trim() || \"https://{region}.tts.speech.microsoft.com/cognitiveservices/v1\";\n  return endpoint.replace(/\\{region\\}/g, encodeURIComponent(config.region));\n}\n\nexport function createSpeechConfig(config: TtsConfig): SpeechConfig {\n  const { outputFormat = DEFAULT_OUTPUT_FORMAT, subscriptionKey } = config;\n\n  const endpoint = new URL(resolveEndpoint(config));\n  const speechConfig = SpeechConfig.fromEndpoint(endpoint, subscriptionKey);\n  speechConfig.speechSynthesisOutputFormat = resolveOutputFormat(outputFormat);\n  return speechConfig;\n}\n","import * as SpeechSDK from \"microsoft-cognitiveservices-speech-sdk\";\n\nexport const DEFAULT_OUTPUT_FORMAT = \"audio-16khz-128kbitrate-mono-mp3\";\n\nconst OUTPUT_FORMATS: Record<string, SpeechSDK.SpeechSynthesisOutputFormat> = {\n  \"raw-8khz-8bit-mono-mulaw\": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz8BitMonoMULaw,\n  \"riff-16khz-16kbps-mono-siren\": SpeechSDK.SpeechSynthesisOutputFormat.Riff16Khz16KbpsMonoSiren,\n  \"audio-16khz-16kbps-mono-siren\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz16KbpsMonoSiren,\n  \"audio-16khz-32kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz32KBitRateMonoMp3,\n  \"audio-16khz-128kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz128KBitRateMonoMp3,\n  \"audio-16khz-64kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz64KBitRateMonoMp3,\n  \"audio-24khz-48kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz48KBitRateMonoMp3,\n  \"audio-24khz-96kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz96KBitRateMonoMp3,\n  \"audio-24khz-160kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz160KBitRateMonoMp3,\n  \"raw-16khz-16bit-mono-truesilk\": SpeechSDK.SpeechSynthesisOutputFormat.Raw16Khz16BitMonoTrueSilk,\n  \"riff-16khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff16Khz16BitMonoPcm,\n  \"riff-8khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz16BitMonoPcm,\n  \"riff-24khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff24Khz16BitMonoPcm,\n  \"riff-8khz-8bit-mono-mulaw\": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz8BitMonoMULaw,\n  \"raw-16khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw16Khz16BitMonoPcm,\n  \"raw-24khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw24Khz16BitMonoPcm,\n  \"raw-8khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz16BitMonoPcm,\n  \"ogg-16khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Ogg16Khz16BitMonoOpus,\n  \"ogg-24khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Ogg24Khz16BitMonoOpus,\n  \"raw-48khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw48Khz16BitMonoPcm,\n  \"riff-48khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff48Khz16BitMonoPcm,\n  \"audio-48khz-96kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio48Khz96KBitRateMonoMp3,\n  \"audio-48khz-192kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio48Khz192KBitRateMonoMp3,\n  \"ogg-48khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Ogg48Khz16BitMonoOpus,\n  \"webm-16khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Webm16Khz16BitMonoOpus,\n  \"webm-24khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Webm24Khz16BitMonoOpus,\n  \"webm-24khz-16bit-24kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Webm24Khz16Bit24KbpsMonoOpus,\n  \"raw-24khz-16bit-mono-truesilk\": SpeechSDK.SpeechSynthesisOutputFormat.Raw24Khz16BitMonoTrueSilk,\n  \"raw-8khz-8bit-mono-alaw\": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz8BitMonoALaw,\n  \"riff-8khz-8bit-mono-alaw\": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz8BitMonoALaw,\n  \"audio-16khz-16bit-32kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz16Bit32KbpsMonoOpus,\n  \"audio-24khz-16bit-48kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz16Bit48KbpsMonoOpus,\n  \"audio-24khz-16bit-24kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz16Bit24KbpsMonoOpus,\n  \"raw-22050hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw22050Hz16BitMonoPcm,\n  \"riff-22050hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff22050Hz16BitMonoPcm,\n  \"raw-44100hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw44100Hz16BitMonoPcm,\n  \"riff-44100hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff44100Hz16BitMonoPcm,\n  \"amr-wb-16000hz\": SpeechSDK.SpeechSynthesisOutputFormat.AmrWb16000Hz,\n  \"g722-16khz-64kbps\": SpeechSDK.SpeechSynthesisOutputFormat.G72216Khz64Kbps,\n};\n\nexport function resolveOutputFormat(outputFormat: string): SpeechSDK.SpeechSynthesisOutputFormat {\n  const resolvedFormat = OUTPUT_FORMATS[outputFormat];\n  if (resolvedFormat === undefined) {\n    throw new Error(`Unsupported Azure Speech output format: ${outputFormat}`);\n  }\n\n  return resolvedFormat;\n}\n","import { synthesizeSpeech } from \"./synthesis.ts\";\nimport type { AzureTtsClientOptions } from \"./types.ts\";\n\nconst ENDPOINT_TEMPLATE = \"https://{region}.tts.speech.microsoft.com/cognitiveservices/v1\";\n\nexport class AzureTtsClient {\n  readonly #options: AzureTtsClientOptions;\n\n  constructor(options: AzureTtsClientOptions) {\n    this.#options = options;\n  }\n\n  async synthesize(ssml: string): Promise<ArrayBuffer> {\n    const { region, subscriptionKey, outputFormat, signal, timeoutMs } = this.#options;\n    const endpoint = this.#options.endpoint?.trim() || ENDPOINT_TEMPLATE.replace(\"{region}\", region);\n    this.#options.logger?.debug?.(\"Using Azure TTS endpoint:\", endpoint);\n\n    const config = { endpoint, region, subscriptionKey, outputFormat, signal, timeoutMs };\n    return synthesizeSpeech(ssml, config);\n  }\n}\n"],"mappings":";;;;;;;;;AAAO,IAAM,gBAAN,cAA4B,MAAM;AAAA,EAMvC,YAAY,QAAgB,YAAoB,cAAsB,WAA0B;AAC9F,UAAM,6BAA6B,MAAM,IAAI,UAAU,EAAE;AACzD,SAAK,OAAO;AACZ,SAAK,SAAS;AACd,SAAK,aAAa;AAClB,SAAK,eAAe;AACpB,SAAK,YAAY;AAAA,EACnB;AACF;AAEO,IAAM,mBAAN,cAA+B,cAAc;AAAA,EAGlD,YAAY,cAAsB;AAChC,UAAM,GAAG,cAAc,cAAc,IAAI;AACzC,SAAK,OAAO;AACZ,SAAK,UAAU,+BAA+B,YAAY;AAC1D,SAAK,eAAe;AAAA,EACtB;AACF;AAEO,SAAS,qBAAqB,OAAkC;AACrE,QAAM,UAAU,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AACrE,SAAO,IAAI,iBAAiB,OAAO;AACrC;;;AC9BA,YAAYA,gBAAe;;;ACA3B,SAAS,oBAAoB;;;ACA7B,YAAY,eAAe;AAEpB,IAAM,wBAAwB;AAErC,IAAM,iBAAwE;AAAA,EAC5E,4BAAsC,sCAA4B;AAAA,EAClE,gCAA0C,sCAA4B;AAAA,EACtE,iCAA2C,sCAA4B;AAAA,EACvE,mCAA6C,sCAA4B;AAAA,EACzE,oCAA8C,sCAA4B;AAAA,EAC1E,mCAA6C,sCAA4B;AAAA,EACzE,mCAA6C,sCAA4B;AAAA,EACzE,mCAA6C,sCAA4B;AAAA,EACzE,oCAA8C,sCAA4B;AAAA,EAC1E,iCAA2C,sCAA4B;AAAA,EACvE,6BAAuC,sCAA4B;AAAA,EACnE,4BAAsC,sCAA4B;AAAA,EAClE,6BAAuC,sCAA4B;AAAA,EACnE,6BAAuC,sCAA4B;AAAA,EACnE,4BAAsC,sCAA4B;AAAA,EAClE,4BAAsC,sCAA4B;AAAA,EAClE,2BAAqC,sCAA4B;AAAA,EACjE,6BAAuC,sCAA4B;AAAA,EACnE,6BAAuC,sCAA4B;AAAA,EACnE,4BAAsC,sCAA4B;AAAA,EAClE,6BAAuC,sCAA4B;AAAA,EACnE,mCAA6C,sCAA4B;AAAA,EACzE,oCAA8C,sCAA4B;AAAA,EAC1E,6BAAuC,sCAA4B;AAAA,EACnE,8BAAwC,sCAA4B;AAAA,EACpE,8BAAwC,sCAA4B;AAAA,EACpE,qCAA+C,sCAA4B;AAAA,EAC3E,iCAA2C,sCAA4B;AAAA,EACvE,2BAAqC,sCAA4B;AAAA,EACjE,4BAAsC,sCAA4B;AAAA,EAClE,sCAAgD,sCAA4B;AAAA,EAC5E,sCAAgD,sCAA4B;AAAA,EAC5E,sCAAgD,sCAA4B;AAAA,EAC5E,8BAAwC,sCAA4B;AAAA,EACpE,+BAAyC,sCAA4B;AAAA,EACrE,8BAAwC,sCAA4B;AAAA,EACpE,+BAAyC,sCAA4B;AAAA,EACrE,kBAA4B,sCAA4B;AAAA,EACxD,qBAA+B,sCAA4B;AAC7D;AAEO,SAAS,oBAAoB,cAA6D;AAC/F,QAAM,iBAAiB,eAAe,YAAY;AAClD,MAAI,mBAAmB,QAAW;AAChC,UAAM,IAAI,MAAM,2CAA2C,YAAY,EAAE;AAAA,EAC3E;AAEA,SAAO;AACT;;;ADjDO,SAAS,gBAAgB,QAA2B;AACzD,QAAM,WAAW,OAAO,UAAU,KAAK,KAAK;AAC5C,SAAO,SAAS,QAAQ,eAAe,mBAAmB,OAAO,MAAM,CAAC;AAC1E;AAEO,SAAS,mBAAmB,QAAiC;AAClE,QAAM,EAAE,eAAe,uBAAuB,gBAAgB,IAAI;AAElE,QAAM,WAAW,IAAI,IAAI,gBAAgB,MAAM,CAAC;AAChD,QAAM,eAAe,aAAa,aAAa,UAAU,eAAe;AACxE,eAAa,8BAA8B,oBAAoB,YAAY;AAC3E,SAAO;AACT;;;ADXA,SAAS,qBAAqB,cAAsC,aAAgD;AAClH,MAAI;AACF,gBAAY,MAAM;AAAA,EACpB,QAAQ;AAAA,EAAC;AAET,MAAI;AACF,iBAAa,MAAM;AAAA,EACrB,QAAQ;AAAA,EAAC;AACX;AAEA,eAAsB,iBAAiB,MAAc,QAAyC;AAC5F,MAAI,OAAO,QAAQ,SAAS;AAC1B,UAAM,qBAAqB,iCAAiC;AAAA,EAC9D;AAEA,QAAM,eAAe,mBAAmB,MAAM;AAC9C,QAAM,cAAc,IAAc,6BAAkB,cAAc,IAAI;AAEtE,SAAO,MAAM,IAAI,QAAqB,CAAC,SAAS,WAAW;AACzD,QAAI,kBAAkB;AACtB,QAAI,UAAU;AACd,QAAI;AACJ,QAAI;AACJ,UAAM,UAAU,MAAM;AACpB,UAAI,QAAS,cAAa,OAAO;AACjC,UAAI,aAAc,QAAO,QAAQ,oBAAoB,SAAS,YAAY;AAAA,IAC5E;AACA,UAAM,iBAAiB,MAAM;AAC3B,UAAI,gBAAiB;AACrB,wBAAkB;AAClB,2BAAqB,cAAc,WAAW;AAAA,IAChD;AACA,UAAM,kBAAkB,CAAC,UAAmB;AAC1C,UAAI,QAAS;AACb,gBAAU;AACV,cAAQ;AACR,qBAAe;AACf,aAAO,qBAAqB,KAAK,CAAC;AAAA,IACpC;AAEA,UAAM,KAAK,CAAC,WAA4C;AACtD,UAAI,QAAS;AACb,YAAM,EAAE,QAAQ,aAAa,IAAI;AACjC,UAAI,WAAqB,wBAAa,4BAA4B;AAChE,cAAM,MAAM,gBAAgB,uCAAuC,MAAM;AACzE,wBAAgB,GAAG;AACnB;AAAA,MACF;AACA,gBAAU;AACV,cAAQ;AACR,qBAAe;AACf,cAAQ,OAAO,SAAS;AAAA,IAC1B;AAEA,QAAI;AACF,UAAI,OAAO,QAAQ;AACjB,uBAAe,MAAM,gBAAgB,iCAAiC;AACtE,eAAO,OAAO,iBAAiB,SAAS,cAAc,EAAE,MAAM,KAAK,CAAC;AAAA,MACtE;AACA,UAAI,OAAO,cAAc,UAAa,OAAO,YAAY,GAAG;AAC1D,kBAAU;AAAA,UACR,MAAM,gBAAgB,oCAAoC,OAAO,SAAS,MAAM;AAAA,UAChF,OAAO;AAAA,QACT;AAAA,MACF;AACA,kBAAY,eAAe,MAAM,IAAI,eAAe;AAAA,IACtD,SAAS,OAAO;AACd,sBAAgB,KAAK;AAAA,IACvB;AAAA,EACF,CAAC;AACH;;;AGxEA,IAAM,oBAAoB;AAH1B;AAKO,IAAM,iBAAN,MAAqB;AAAA,EAG1B,YAAY,SAAgC;AAF5C,uBAAS;AAGP,uBAAK,UAAW;AAAA,EAClB;AAAA,EAEA,MAAM,WAAW,MAAoC;AACnD,UAAM,EAAE,QAAQ,iBAAiB,cAAc,QAAQ,UAAU,IAAI,mBAAK;AAC1E,UAAM,WAAW,mBAAK,UAAS,UAAU,KAAK,KAAK,kBAAkB,QAAQ,YAAY,MAAM;AAC/F,uBAAK,UAAS,QAAQ,QAAQ,6BAA6B,QAAQ;AAEnE,UAAM,SAAS,EAAE,UAAU,QAAQ,iBAAiB,cAAc,QAAQ,UAAU;AACpF,WAAO,iBAAiB,MAAM,MAAM;AAAA,EACtC;AACF;AAdW;","names":["SpeechSDK"]}
+{"version":3,"sources":["../src/errors.ts","../src/synthesis.ts","../src/speechConfig.ts","../src/outputFormats.ts","../src/client.ts","../src/voiceCatalog.ts"],"sourcesContent":["export class AzureTtsError extends Error {\n  readonly status: number;\n  readonly statusText: string;\n  readonly responseBody: string;\n  readonly requestId: string | null;\n\n  constructor(status: number, statusText: string, responseBody: string, requestId: string | null) {\n    super(`Azure TTS request failed: ${status} ${statusText}`);\n    this.name = \"AzureTtsError\";\n    this.status = status;\n    this.statusText = statusText;\n    this.responseBody = responseBody;\n    this.requestId = requestId;\n  }\n}\n\nexport class AzureTtsSdkError extends AzureTtsError {\n  readonly errorDetails: string;\n\n  constructor(errorDetails: string) {\n    super(0, \"Speech SDK\", errorDetails, null);\n    this.name = \"AzureTtsSdkError\";\n    this.message = `Azure TTS synthesis failed: ${errorDetails}`;\n    this.errorDetails = errorDetails;\n  }\n}\n\nexport function createSpeechSdkError(error: unknown): AzureTtsSdkError {\n  const message = error instanceof Error ? error.message : String(error);\n  return new AzureTtsSdkError(message);\n}\n","import * as SpeechSDK from \"microsoft-cognitiveservices-speech-sdk\";\nimport { createSpeechSdkError } from \"./errors.ts\";\nimport { createSpeechConfig } from \"./speechConfig.ts\";\nimport type { SsmlSynthesisResult, TtsConfig } from \"./types.ts\";\n\nfunction closeSpeechResources(speechConfig: SpeechSDK.SpeechConfig, synthesizer: SpeechSDK.SpeechSynthesizer): void {\n  try {\n    synthesizer.close();\n  } catch {}\n\n  try {\n    speechConfig.close();\n  } catch {}\n}\n\nconst ticksToMilliseconds = (ticks: number): number => Math.max(0, ticks) / 10_000;\n\nexport async function synthesizeSsml(ssml: string, config: TtsConfig): Promise<SsmlSynthesisResult> {\n  if (config.signal?.aborted) {\n    throw createSpeechSdkError(\"Speech synthesis was cancelled.\");\n  }\n\n  const speechConfig = createSpeechConfig(config);\n  const synthesizer = new SpeechSDK.SpeechSynthesizer(speechConfig, null);\n\n  return await new Promise<SsmlSynthesisResult>((resolve, reject) => {\n    let resourcesClosed = false;\n    let settled = false;\n    let timeout: ReturnType<typeof setTimeout> | undefined;\n    let abortHandler: (() => void) | undefined;\n    const cleanup = () => {\n      if (timeout) clearTimeout(timeout);\n      if (abortHandler) config.signal?.removeEventListener(\"abort\", abortHandler);\n    };\n    const closeResources = () => {\n      if (resourcesClosed) return;\n      resourcesClosed = true;\n      closeSpeechResources(speechConfig, synthesizer);\n    };\n    const rejectWithError = (error: unknown) => {\n      if (settled) return;\n      settled = true;\n      cleanup();\n      closeResources();\n      reject(createSpeechSdkError(error));\n    };\n\n    const boundaries: SsmlSynthesisResult[\"boundaries\"] = [];\n    const visemes: SsmlSynthesisResult[\"visemes\"] = [];\n    const bookmarks: SsmlSynthesisResult[\"bookmarks\"] = [];\n    synthesizer.wordBoundary = (_sender, event) => {\n      boundaries.push({\n        text: event.text,\n        audioOffsetMs: ticksToMilliseconds(event.audioOffset),\n        durationMs: ticksToMilliseconds(event.duration),\n      });\n    };\n    synthesizer.visemeReceived = (_sender, event) => {\n      visemes.push({ visemeId: event.visemeId, audioOffsetMs: ticksToMilliseconds(event.audioOffset) });\n    };\n    synthesizer.bookmarkReached = (_sender, event) => {\n      bookmarks.push({ name: event.text, audioOffsetMs: ticksToMilliseconds(event.audioOffset) });\n    };\n\n    const cb = (result: SpeechSDK.SpeechSynthesisResult) => {\n      if (settled) return;\n      const { reason, errorDetails } = result;\n      if (reason !== SpeechSDK.ResultReason.SynthesizingAudioCompleted) {\n        const err = errorDetails || `Speech synthesis failed with reason ${reason}.`;\n        rejectWithError(err);\n        return;\n      }\n      settled = true;\n      cleanup();\n      closeResources();\n      const eventDurationMs = Math.max(\n        0,\n        ...(boundaries ?? []).map((boundary) => boundary.audioOffsetMs + boundary.durationMs),\n        ...(visemes ?? []).map((viseme) => viseme.audioOffsetMs),\n        ...(bookmarks ?? []).map((bookmark) => bookmark.audioOffsetMs),\n      );\n      const durationMs = result.audioDuration ? ticksToMilliseconds(result.audioDuration) : eventDurationMs;\n      resolve({\n        audioData: result.audioData,\n        durationMs,\n        ...(boundaries.length > 0 ? { boundaries, wordBoundary: boundaries, wordBoundaries: boundaries } : {}),\n        ...(visemes.length > 0 ? { visemes } : {}),\n        ...(bookmarks.length > 0 ? { bookmarks } : {}),\n      });\n    };\n\n    try {\n      if (config.signal) {\n        abortHandler = () => rejectWithError(\"Speech synthesis was cancelled.\");\n        config.signal.addEventListener(\"abort\", abortHandler, { once: true });\n      }\n      if (config.timeoutMs !== undefined && config.timeoutMs > 0) {\n        timeout = setTimeout(\n          () => rejectWithError(`Speech synthesis timed out after ${config.timeoutMs} ms.`),\n          config.timeoutMs,\n        );\n      }\n      synthesizer.speakSsmlAsync(ssml, cb, rejectWithError);\n    } catch (error) {\n      rejectWithError(error);\n    }\n  });\n}\n\n/** Backward-compatible audio-only synthesis helper. */\nexport async function synthesizeSpeech(ssml: string, config: TtsConfig): Promise<ArrayBuffer> {\n  return (await synthesizeSsml(ssml, config)).audioData;\n}\n","import { SpeechConfig } from \"microsoft-cognitiveservices-speech-sdk\";\nimport { DEFAULT_OUTPUT_FORMAT, resolveOutputFormat } from \"./outputFormats.ts\";\nimport type { TtsConfig } from \"./types.ts\";\n\nexport function resolveEndpoint(config: TtsConfig): string {\n  const endpoint = config.endpoint?.trim() || \"https://{region}.tts.speech.microsoft.com/cognitiveservices/v1\";\n  return endpoint.replace(/\\{region\\}/g, encodeURIComponent(config.region));\n}\n\nexport function createSpeechConfig(config: TtsConfig): SpeechConfig {\n  const { outputFormat = DEFAULT_OUTPUT_FORMAT, subscriptionKey } = config;\n\n  const endpoint = new URL(resolveEndpoint(config));\n  const speechConfig = SpeechConfig.fromEndpoint(endpoint, subscriptionKey);\n  speechConfig.speechSynthesisOutputFormat = resolveOutputFormat(outputFormat);\n  return speechConfig;\n}\n","import * as SpeechSDK from \"microsoft-cognitiveservices-speech-sdk\";\n\nexport const DEFAULT_OUTPUT_FORMAT = \"audio-16khz-128kbitrate-mono-mp3\";\n\nconst OUTPUT_FORMATS: Record<string, SpeechSDK.SpeechSynthesisOutputFormat> = {\n  \"raw-8khz-8bit-mono-mulaw\": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz8BitMonoMULaw,\n  \"riff-16khz-16kbps-mono-siren\": SpeechSDK.SpeechSynthesisOutputFormat.Riff16Khz16KbpsMonoSiren,\n  \"audio-16khz-16kbps-mono-siren\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz16KbpsMonoSiren,\n  \"audio-16khz-32kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz32KBitRateMonoMp3,\n  \"audio-16khz-128kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz128KBitRateMonoMp3,\n  \"audio-16khz-64kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz64KBitRateMonoMp3,\n  \"audio-24khz-48kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz48KBitRateMonoMp3,\n  \"audio-24khz-96kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz96KBitRateMonoMp3,\n  \"audio-24khz-160kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz160KBitRateMonoMp3,\n  \"raw-16khz-16bit-mono-truesilk\": SpeechSDK.SpeechSynthesisOutputFormat.Raw16Khz16BitMonoTrueSilk,\n  \"riff-16khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff16Khz16BitMonoPcm,\n  \"riff-8khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz16BitMonoPcm,\n  \"riff-24khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff24Khz16BitMonoPcm,\n  \"riff-8khz-8bit-mono-mulaw\": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz8BitMonoMULaw,\n  \"raw-16khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw16Khz16BitMonoPcm,\n  \"raw-24khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw24Khz16BitMonoPcm,\n  \"raw-8khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz16BitMonoPcm,\n  \"ogg-16khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Ogg16Khz16BitMonoOpus,\n  \"ogg-24khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Ogg24Khz16BitMonoOpus,\n  \"raw-48khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw48Khz16BitMonoPcm,\n  \"riff-48khz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff48Khz16BitMonoPcm,\n  \"audio-48khz-96kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio48Khz96KBitRateMonoMp3,\n  \"audio-48khz-192kbitrate-mono-mp3\": SpeechSDK.SpeechSynthesisOutputFormat.Audio48Khz192KBitRateMonoMp3,\n  \"ogg-48khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Ogg48Khz16BitMonoOpus,\n  \"webm-16khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Webm16Khz16BitMonoOpus,\n  \"webm-24khz-16bit-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Webm24Khz16BitMonoOpus,\n  \"webm-24khz-16bit-24kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Webm24Khz16Bit24KbpsMonoOpus,\n  \"raw-24khz-16bit-mono-truesilk\": SpeechSDK.SpeechSynthesisOutputFormat.Raw24Khz16BitMonoTrueSilk,\n  \"raw-8khz-8bit-mono-alaw\": SpeechSDK.SpeechSynthesisOutputFormat.Raw8Khz8BitMonoALaw,\n  \"riff-8khz-8bit-mono-alaw\": SpeechSDK.SpeechSynthesisOutputFormat.Riff8Khz8BitMonoALaw,\n  \"audio-16khz-16bit-32kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Audio16Khz16Bit32KbpsMonoOpus,\n  \"audio-24khz-16bit-48kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz16Bit48KbpsMonoOpus,\n  \"audio-24khz-16bit-24kbps-mono-opus\": SpeechSDK.SpeechSynthesisOutputFormat.Audio24Khz16Bit24KbpsMonoOpus,\n  \"raw-22050hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw22050Hz16BitMonoPcm,\n  \"riff-22050hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff22050Hz16BitMonoPcm,\n  \"raw-44100hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Raw44100Hz16BitMonoPcm,\n  \"riff-44100hz-16bit-mono-pcm\": SpeechSDK.SpeechSynthesisOutputFormat.Riff44100Hz16BitMonoPcm,\n  \"amr-wb-16000hz\": SpeechSDK.SpeechSynthesisOutputFormat.AmrWb16000Hz,\n  \"g722-16khz-64kbps\": SpeechSDK.SpeechSynthesisOutputFormat.G72216Khz64Kbps,\n};\n\nexport function resolveOutputFormat(outputFormat: string): SpeechSDK.SpeechSynthesisOutputFormat {\n  const resolvedFormat = OUTPUT_FORMATS[outputFormat];\n  if (resolvedFormat === undefined) {\n    throw new Error(`Unsupported Azure Speech output format: ${outputFormat}`);\n  }\n\n  return resolvedFormat;\n}\n","import { synthesizeSpeech, synthesizeSsml } from \"./synthesis.ts\";\nimport type { AzureTtsClientOptions, SsmlSynthesisResult } from \"./types.ts\";\n\nconst ENDPOINT_TEMPLATE = \"https://{region}.tts.speech.microsoft.com/cognitiveservices/v1\";\n\nexport class AzureTtsClient {\n  readonly #options: AzureTtsClientOptions;\n\n  constructor(options: AzureTtsClientOptions) {\n    this.#options = options;\n  }\n\n  async synthesize(ssml: string): Promise<ArrayBuffer> {\n    const { region, subscriptionKey, outputFormat, signal, timeoutMs } = this.#options;\n    const endpoint = this.#options.endpoint?.trim() || ENDPOINT_TEMPLATE.replace(\"{region}\", region);\n    this.#options.logger?.debug?.(\"Using Azure TTS endpoint:\", endpoint);\n\n    const config = { endpoint, region, subscriptionKey, outputFormat, signal, timeoutMs };\n    return synthesizeSpeech(ssml, config);\n  }\n\n  async synthesizeSsml(ssml: string): Promise<SsmlSynthesisResult> {\n    const { region, subscriptionKey, outputFormat, signal, timeoutMs } = this.#options;\n    const endpoint = this.#options.endpoint?.trim() || ENDPOINT_TEMPLATE.replace(\"{region}\", region);\n    this.#options.logger?.debug?.(\"Using Azure TTS endpoint:\", endpoint);\n\n    return synthesizeSsml(ssml, { endpoint, region, subscriptionKey, outputFormat, signal, timeoutMs });\n  }\n}\n","const AZURE_VOICE_API_VERSION = \"2025-10-01\";\n\nexport interface FetchAzureVoiceCatalogOptions {\n  apiKey: string;\n  region: string | string[];\n}\n\nexport interface AzureVoiceCatalogVoice {\n  name: string;\n  locale: string;\n  secondaryLocales?: readonly string[];\n  styles?: readonly string[];\n  regions: readonly string[];\n  status?: \"ga\" | \"preview\" | \"deprecated\";\n}\n\nexport interface FetchedAzureVoiceCatalogMetadata {\n  voiceCount: number;\n  generatedAt: string;\n  apiVersion: string;\n  regions: readonly string[];\n}\n\nexport interface AzureVoiceCatalog {\n  voices: readonly AzureVoiceCatalogVoice[];\n  metadata: FetchedAzureVoiceCatalogMetadata;\n}\n\ninterface AzureVoiceApiRecord {\n  Locale?: unknown;\n  Name?: unknown;\n  SecondaryLocaleList?: unknown;\n  ShortName?: unknown;\n  Status?: unknown;\n  StyleList?: unknown;\n}\n\nfunction stringValue(value: unknown): string | undefined {\n  return typeof value === \"string\" && value.trim() ? value.trim() : undefined;\n}\n\nfunction stringList(value: unknown): string[] {\n  if (!Array.isArray(value)) return [];\n  return [...new Set(value.map(stringValue).filter((item): item is string => item !== undefined))];\n}\n\nfunction normalizeStatus(value: unknown): AzureVoiceCatalogVoice[\"status\"] {\n  const status = stringValue(value)?.toLowerCase();\n  if (status === \"preview\" || status === \"deprecated\" || status === \"ga\") return status;\n  return undefined;\n}\n\nfunction normalizeRegions(region: string | string[]): string[] {\n  const regions = Array.isArray(region) ? region : [region];\n  const result = [...new Set(regions.map((item) => item.trim()).filter(Boolean))];\n  if (result.length === 0) throw new TypeError(\"At least one Azure Speech region is required.\");\n  return result;\n}\n\nasync function fetchRegionVoices(region: string, apiKey: string): Promise<AzureVoiceApiRecord[]> {\n  const endpoint = `https://${encodeURIComponent(region)}.tts.speech.microsoft.com/cognitiveservices/voices/list`;\n  const response = await fetch(endpoint, {\n    headers: {\n      Accept: \"application/json\",\n      \"Ocp-Apim-Subscription-Key\": apiKey,\n    },\n  });\n  if (!response.ok) {\n    throw new Error(`Azure List Voices API request failed for region \"${region}\" with HTTP ${response.status}.`);\n  }\n  const payload: unknown = await response.json();\n  if (!Array.isArray(payload)) throw new Error(`Azure List Voices API returned an invalid response for \"${region}\".`);\n  return payload.filter((item): item is AzureVoiceApiRecord => Boolean(item && typeof item === \"object\"));\n}\n\n/** Fetches and deduplicates the current Azure Speech voice catalog for one or more regions. */\nexport async function fetchAzureVoiceCatalog(options: FetchAzureVoiceCatalogOptions): Promise<AzureVoiceCatalog> {\n  if (!options || typeof options.apiKey !== \"string\" || !options.apiKey.trim())\n    throw new TypeError(\"An Azure Speech API key is required.\");\n  const regions = normalizeRegions(options.region);\n  const payloads = await Promise.all(regions.map((region) => fetchRegionVoices(region, options.apiKey)));\n  const voices = new Map<string, AzureVoiceCatalogVoice>();\n\n  for (let regionIndex = 0; regionIndex < payloads.length; regionIndex += 1) {\n    const region = regions[regionIndex];\n    for (const record of payloads[regionIndex]) {\n      const name = stringValue(record.ShortName) ?? stringValue(record.Name);\n      const locale = stringValue(record.Locale);\n      if (!name || !locale) continue;\n      const key = name.toLowerCase();\n      const existing = voices.get(key);\n      const secondaryLocales = stringList(record.SecondaryLocaleList);\n      const styles = stringList(record.StyleList);\n      const status = normalizeStatus(record.Status);\n      const merged: AzureVoiceCatalogVoice = {\n        name: existing?.name ?? name,\n        locale: existing?.locale ?? locale,\n        regions: [...new Set([...(existing?.regions ?? []), region])],\n      };\n      const mergedSecondaryLocales = [...new Set([...(existing?.secondaryLocales ?? []), ...secondaryLocales])];\n      if (mergedSecondaryLocales.length > 0) merged.secondaryLocales = mergedSecondaryLocales;\n      const mergedStyles = [...new Set([...(existing?.styles ?? []), ...styles])];\n      if (mergedStyles.length > 0) merged.styles = mergedStyles;\n      if (status) merged.status = status;\n      else if (existing?.status) merged.status = existing.status;\n      voices.set(key, merged);\n    }\n  }\n\n  const sortedVoices = [...voices.values()].sort((first, second) => first.name.localeCompare(second.name));\n  return {\n    voices: sortedVoices,\n    metadata: {\n      voiceCount: sortedVoices.length,\n      generatedAt: new Date().toISOString(),\n      apiVersion: AZURE_VOICE_API_VERSION,\n      regions,\n    },\n  };\n}\n"],"mappings":";;;;;;;;;AAAO,IAAM,gBAAN,cAA4B,MAAM;AAAA,EAMvC,YAAY,QAAgB,YAAoB,cAAsB,WAA0B;AAC9F,UAAM,6BAA6B,MAAM,IAAI,UAAU,EAAE;AACzD,SAAK,OAAO;AACZ,SAAK,SAAS;AACd,SAAK,aAAa;AAClB,SAAK,eAAe;AACpB,SAAK,YAAY;AAAA,EACnB;AACF;AAEO,IAAM,mBAAN,cAA+B,cAAc;AAAA,EAGlD,YAAY,cAAsB;AAChC,UAAM,GAAG,cAAc,cAAc,IAAI;AACzC,SAAK,OAAO;AACZ,SAAK,UAAU,+BAA+B,YAAY;AAC1D,SAAK,eAAe;AAAA,EACtB;AACF;AAEO,SAAS,qBAAqB,OAAkC;AACrE,QAAM,UAAU,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AACrE,SAAO,IAAI,iBAAiB,OAAO;AACrC;;;AC9BA,YAAYA,gBAAe;;;ACA3B,SAAS,oBAAoB;;;ACA7B,YAAY,eAAe;AAEpB,IAAM,wBAAwB;AAErC,IAAM,iBAAwE;AAAA,EAC5E,4BAAsC,sCAA4B;AAAA,EAClE,gCAA0C,sCAA4B;AAAA,EACtE,iCAA2C,sCAA4B;AAAA,EACvE,mCAA6C,sCAA4B;AAAA,EACzE,oCAA8C,sCAA4B;AAAA,EAC1E,mCAA6C,sCAA4B;AAAA,EACzE,mCAA6C,sCAA4B;AAAA,EACzE,mCAA6C,sCAA4B;AAAA,EACzE,oCAA8C,sCAA4B;AAAA,EAC1E,iCAA2C,sCAA4B;AAAA,EACvE,6BAAuC,sCAA4B;AAAA,EACnE,4BAAsC,sCAA4B;AAAA,EAClE,6BAAuC,sCAA4B;AAAA,EACnE,6BAAuC,sCAA4B;AAAA,EACnE,4BAAsC,sCAA4B;AAAA,EAClE,4BAAsC,sCAA4B;AAAA,EAClE,2BAAqC,sCAA4B;AAAA,EACjE,6BAAuC,sCAA4B;AAAA,EACnE,6BAAuC,sCAA4B;AAAA,EACnE,4BAAsC,sCAA4B;AAAA,EAClE,6BAAuC,sCAA4B;AAAA,EACnE,mCAA6C,sCAA4B;AAAA,EACzE,oCAA8C,sCAA4B;AAAA,EAC1E,6BAAuC,sCAA4B;AAAA,EACnE,8BAAwC,sCAA4B;AAAA,EACpE,8BAAwC,sCAA4B;AAAA,EACpE,qCAA+C,sCAA4B;AAAA,EAC3E,iCAA2C,sCAA4B;AAAA,EACvE,2BAAqC,sCAA4B;AAAA,EACjE,4BAAsC,sCAA4B;AAAA,EAClE,sCAAgD,sCAA4B;AAAA,EAC5E,sCAAgD,sCAA4B;AAAA,EAC5E,sCAAgD,sCAA4B;AAAA,EAC5E,8BAAwC,sCAA4B;AAAA,EACpE,+BAAyC,sCAA4B;AAAA,EACrE,8BAAwC,sCAA4B;AAAA,EACpE,+BAAyC,sCAA4B;AAAA,EACrE,kBAA4B,sCAA4B;AAAA,EACxD,qBAA+B,sCAA4B;AAC7D;AAEO,SAAS,oBAAoB,cAA6D;AAC/F,QAAM,iBAAiB,eAAe,YAAY;AAClD,MAAI,mBAAmB,QAAW;AAChC,UAAM,IAAI,MAAM,2CAA2C,YAAY,EAAE;AAAA,EAC3E;AAEA,SAAO;AACT;;;ADjDO,SAAS,gBAAgB,QAA2B;AACzD,QAAM,WAAW,OAAO,UAAU,KAAK,KAAK;AAC5C,SAAO,SAAS,QAAQ,eAAe,mBAAmB,OAAO,MAAM,CAAC;AAC1E;AAEO,SAAS,mBAAmB,QAAiC;AAClE,QAAM,EAAE,eAAe,uBAAuB,gBAAgB,IAAI;AAElE,QAAM,WAAW,IAAI,IAAI,gBAAgB,MAAM,CAAC;AAChD,QAAM,eAAe,aAAa,aAAa,UAAU,eAAe;AACxE,eAAa,8BAA8B,oBAAoB,YAAY;AAC3E,SAAO;AACT;;;ADXA,SAAS,qBAAqB,cAAsC,aAAgD;AAClH,MAAI;AACF,gBAAY,MAAM;AAAA,EACpB,QAAQ;AAAA,EAAC;AAET,MAAI;AACF,iBAAa,MAAM;AAAA,EACrB,QAAQ;AAAA,EAAC;AACX;AAEA,IAAM,sBAAsB,CAAC,UAA0B,KAAK,IAAI,GAAG,KAAK,IAAI;AAE5E,eAAsB,eAAe,MAAc,QAAiD;AAClG,MAAI,OAAO,QAAQ,SAAS;AAC1B,UAAM,qBAAqB,iCAAiC;AAAA,EAC9D;AAEA,QAAM,eAAe,mBAAmB,MAAM;AAC9C,QAAM,cAAc,IAAc,6BAAkB,cAAc,IAAI;AAEtE,SAAO,MAAM,IAAI,QAA6B,CAAC,SAAS,WAAW;AACjE,QAAI,kBAAkB;AACtB,QAAI,UAAU;AACd,QAAI;AACJ,QAAI;AACJ,UAAM,UAAU,MAAM;AACpB,UAAI,QAAS,cAAa,OAAO;AACjC,UAAI,aAAc,QAAO,QAAQ,oBAAoB,SAAS,YAAY;AAAA,IAC5E;AACA,UAAM,iBAAiB,MAAM;AAC3B,UAAI,gBAAiB;AACrB,wBAAkB;AAClB,2BAAqB,cAAc,WAAW;AAAA,IAChD;AACA,UAAM,kBAAkB,CAAC,UAAmB;AAC1C,UAAI,QAAS;AACb,gBAAU;AACV,cAAQ;AACR,qBAAe;AACf,aAAO,qBAAqB,KAAK,CAAC;AAAA,IACpC;AAEA,UAAM,aAAgD,CAAC;AACvD,UAAM,UAA0C,CAAC;AACjD,UAAM,YAA8C,CAAC;AACrD,gBAAY,eAAe,CAAC,SAAS,UAAU;AAC7C,iBAAW,KAAK;AAAA,QACd,MAAM,MAAM;AAAA,QACZ,eAAe,oBAAoB,MAAM,WAAW;AAAA,QACpD,YAAY,oBAAoB,MAAM,QAAQ;AAAA,MAChD,CAAC;AAAA,IACH;AACA,gBAAY,iBAAiB,CAAC,SAAS,UAAU;AAC/C,cAAQ,KAAK,EAAE,UAAU,MAAM,UAAU,eAAe,oBAAoB,MAAM,WAAW,EAAE,CAAC;AAAA,IAClG;AACA,gBAAY,kBAAkB,CAAC,SAAS,UAAU;AAChD,gBAAU,KAAK,EAAE,MAAM,MAAM,MAAM,eAAe,oBAAoB,MAAM,WAAW,EAAE,CAAC;AAAA,IAC5F;AAEA,UAAM,KAAK,CAAC,WAA4C;AACtD,UAAI,QAAS;AACb,YAAM,EAAE,QAAQ,aAAa,IAAI;AACjC,UAAI,WAAqB,wBAAa,4BAA4B;AAChE,cAAM,MAAM,gBAAgB,uCAAuC,MAAM;AACzE,wBAAgB,GAAG;AACnB;AAAA,MACF;AACA,gBAAU;AACV,cAAQ;AACR,qBAAe;AACf,YAAM,kBAAkB,KAAK;AAAA,QAC3B;AAAA,QACA,IAAI,cAAc,CAAC,GAAG,IAAI,CAAC,aAAa,SAAS,gBAAgB,SAAS,UAAU;AAAA,QACpF,IAAI,WAAW,CAAC,GAAG,IAAI,CAAC,WAAW,OAAO,aAAa;AAAA,QACvD,IAAI,aAAa,CAAC,GAAG,IAAI,CAAC,aAAa,SAAS,aAAa;AAAA,MAC/D;AACA,YAAM,aAAa,OAAO,gBAAgB,oBAAoB,OAAO,aAAa,IAAI;AACtF,cAAQ;AAAA,QACN,WAAW,OAAO;AAAA,QAClB;AAAA,QACA,GAAI,WAAW,SAAS,IAAI,EAAE,YAAY,cAAc,YAAY,gBAAgB,WAAW,IAAI,CAAC;AAAA,QACpG,GAAI,QAAQ,SAAS,IAAI,EAAE,QAAQ,IAAI,CAAC;AAAA,QACxC,GAAI,UAAU,SAAS,IAAI,EAAE,UAAU,IAAI,CAAC;AAAA,MAC9C,CAAC;AAAA,IACH;AAEA,QAAI;AACF,UAAI,OAAO,QAAQ;AACjB,uBAAe,MAAM,gBAAgB,iCAAiC;AACtE,eAAO,OAAO,iBAAiB,SAAS,cAAc,EAAE,MAAM,KAAK,CAAC;AAAA,MACtE;AACA,UAAI,OAAO,cAAc,UAAa,OAAO,YAAY,GAAG;AAC1D,kBAAU;AAAA,UACR,MAAM,gBAAgB,oCAAoC,OAAO,SAAS,MAAM;AAAA,UAChF,OAAO;AAAA,QACT;AAAA,MACF;AACA,kBAAY,eAAe,MAAM,IAAI,eAAe;AAAA,IACtD,SAAS,OAAO;AACd,sBAAgB,KAAK;AAAA,IACvB;AAAA,EACF,CAAC;AACH;AAGA,eAAsB,iBAAiB,MAAc,QAAyC;AAC5F,UAAQ,MAAM,eAAe,MAAM,MAAM,GAAG;AAC9C;;;AG7GA,IAAM,oBAAoB;AAH1B;AAKO,IAAM,iBAAN,MAAqB;AAAA,EAG1B,YAAY,SAAgC;AAF5C,uBAAS;AAGP,uBAAK,UAAW;AAAA,EAClB;AAAA,EAEA,MAAM,WAAW,MAAoC;AACnD,UAAM,EAAE,QAAQ,iBAAiB,cAAc,QAAQ,UAAU,IAAI,mBAAK;AAC1E,UAAM,WAAW,mBAAK,UAAS,UAAU,KAAK,KAAK,kBAAkB,QAAQ,YAAY,MAAM;AAC/F,uBAAK,UAAS,QAAQ,QAAQ,6BAA6B,QAAQ;AAEnE,UAAM,SAAS,EAAE,UAAU,QAAQ,iBAAiB,cAAc,QAAQ,UAAU;AACpF,WAAO,iBAAiB,MAAM,MAAM;AAAA,EACtC;AAAA,EAEA,MAAM,eAAe,MAA4C;AAC/D,UAAM,EAAE,QAAQ,iBAAiB,cAAc,QAAQ,UAAU,IAAI,mBAAK;AAC1E,UAAM,WAAW,mBAAK,UAAS,UAAU,KAAK,KAAK,kBAAkB,QAAQ,YAAY,MAAM;AAC/F,uBAAK,UAAS,QAAQ,QAAQ,6BAA6B,QAAQ;AAEnE,WAAO,eAAe,MAAM,EAAE,UAAU,QAAQ,iBAAiB,cAAc,QAAQ,UAAU,CAAC;AAAA,EACpG;AACF;AAtBW;;;ACNX,IAAM,0BAA0B;AAqChC,SAAS,YAAY,OAAoC;AACvD,SAAO,OAAO,UAAU,YAAY,MAAM,KAAK,IAAI,MAAM,KAAK,IAAI;AACpE;AAEA,SAAS,WAAW,OAA0B;AAC5C,MAAI,CAAC,MAAM,QAAQ,KAAK,EAAG,QAAO,CAAC;AACnC,SAAO,CAAC,GAAG,IAAI,IAAI,MAAM,IAAI,WAAW,EAAE,OAAO,CAAC,SAAyB,SAAS,MAAS,CAAC,CAAC;AACjG;AAEA,SAAS,gBAAgB,OAAkD;AACzE,QAAM,SAAS,YAAY,KAAK,GAAG,YAAY;AAC/C,MAAI,WAAW,aAAa,WAAW,gBAAgB,WAAW,KAAM,QAAO;AAC/E,SAAO;AACT;AAEA,SAAS,iBAAiB,QAAqC;AAC7D,QAAM,UAAU,MAAM,QAAQ,MAAM,IAAI,SAAS,CAAC,MAAM;AACxD,QAAM,SAAS,CAAC,GAAG,IAAI,IAAI,QAAQ,IAAI,CAAC,SAAS,KAAK,KAAK,CAAC,EAAE,OAAO,OAAO,CAAC,CAAC;AAC9E,MAAI,OAAO,WAAW,EAAG,OAAM,IAAI,UAAU,+CAA+C;AAC5F,SAAO;AACT;AAEA,eAAe,kBAAkB,QAAgB,QAAgD;AAC/F,QAAM,WAAW,WAAW,mBAAmB,MAAM,CAAC;AACtD,QAAM,WAAW,MAAM,MAAM,UAAU;AAAA,IACrC,SAAS;AAAA,MACP,QAAQ;AAAA,MACR,6BAA6B;AAAA,IAC/B;AAAA,EACF,CAAC;AACD,MAAI,CAAC,SAAS,IAAI;AAChB,UAAM,IAAI,MAAM,oDAAoD,MAAM,eAAe,SAAS,MAAM,GAAG;AAAA,EAC7G;AACA,QAAM,UAAmB,MAAM,SAAS,KAAK;AAC7C,MAAI,CAAC,MAAM,QAAQ,OAAO,EAAG,OAAM,IAAI,MAAM,2DAA2D,MAAM,IAAI;AAClH,SAAO,QAAQ,OAAO,CAAC,SAAsC,QAAQ,QAAQ,OAAO,SAAS,QAAQ,CAAC;AACxG;AAGA,eAAsB,uBAAuB,SAAoE;AAC/G,MAAI,CAAC,WAAW,OAAO,QAAQ,WAAW,YAAY,CAAC,QAAQ,OAAO,KAAK;AACzE,UAAM,IAAI,UAAU,sCAAsC;AAC5D,QAAM,UAAU,iBAAiB,QAAQ,MAAM;AAC/C,QAAM,WAAW,MAAM,QAAQ,IAAI,QAAQ,IAAI,CAAC,WAAW,kBAAkB,QAAQ,QAAQ,MAAM,CAAC,CAAC;AACrG,QAAM,SAAS,oBAAI,IAAoC;AAEvD,WAAS,cAAc,GAAG,cAAc,SAAS,QAAQ,eAAe,GAAG;AACzE,UAAM,SAAS,QAAQ,WAAW;AAClC,eAAW,UAAU,SAAS,WAAW,GAAG;AAC1C,YAAM,OAAO,YAAY,OAAO,SAAS,KAAK,YAAY,OAAO,IAAI;AACrE,YAAM,SAAS,YAAY,OAAO,MAAM;AACxC,UAAI,CAAC,QAAQ,CAAC,OAAQ;AACtB,YAAM,MAAM,KAAK,YAAY;AAC7B,YAAM,WAAW,OAAO,IAAI,GAAG;AAC/B,YAAM,mBAAmB,WAAW,OAAO,mBAAmB;AAC9D,YAAM,SAAS,WAAW,OAAO,SAAS;AAC1C,YAAM,SAAS,gBAAgB,OAAO,MAAM;AAC5C,YAAM,SAAiC;AAAA,QACrC,MAAM,UAAU,QAAQ;AAAA,QACxB,QAAQ,UAAU,UAAU;AAAA,QAC5B,SAAS,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAI,UAAU,WAAW,CAAC,GAAI,MAAM,CAAC,CAAC;AAAA,MAC9D;AACA,YAAM,yBAAyB,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAI,UAAU,oBAAoB,CAAC,GAAI,GAAG,gBAAgB,CAAC,CAAC;AACxG,UAAI,uBAAuB,SAAS,EAAG,QAAO,mBAAmB;AACjE,YAAM,eAAe,CAAC,GAAG,oBAAI,IAAI,CAAC,GAAI,UAAU,UAAU,CAAC,GAAI,GAAG,MAAM,CAAC,CAAC;AAC1E,UAAI,aAAa,SAAS,EAAG,QAAO,SAAS;AAC7C,UAAI,OAAQ,QAAO,SAAS;AAAA,eACnB,UAAU,OAAQ,QAAO,SAAS,SAAS;AACpD,aAAO,IAAI,KAAK,MAAM;AAAA,IACxB;AAAA,EACF;AAEA,QAAM,eAAe,CAAC,GAAG,OAAO,OAAO,CAAC,EAAE,KAAK,CAAC,OAAO,WAAW,MAAM,KAAK,cAAc,OAAO,IAAI,CAAC;AACvG,SAAO;AAAA,IACL,QAAQ;AAAA,IACR,UAAU;AAAA,MACR,YAAY,aAAa;AAAA,MACzB,cAAa,oBAAI,KAAK,GAAE,YAAY;AAAA,MACpC,YAAY;AAAA,MACZ;AAAA,IACF;AAAA,EACF;AACF;","names":["SpeechSDK"]}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ssml-builder-js/azure-tts-client",
-  "version": "2.10.0",
+  "version": "2.12.0",
   "description": "Azure Text-to-Speech client using the Microsoft Speech SDK",
   "license": "MIT",
   "repository": {

package/src/client.ts

@@ -1,5 +1,5 @@
-import { synthesizeSpeech } from "./synthesis.ts";
-import type { AzureTtsClientOptions } from "./types.ts";
+import { synthesizeSpeech, synthesizeSsml } from "./synthesis.ts";
+import type { AzureTtsClientOptions, SsmlSynthesisResult } from "./types.ts";
 
 const ENDPOINT_TEMPLATE = "https://{region}.tts.speech.microsoft.com/cognitiveservices/v1";
 
@@ -18,4 +18,12 @@ export class AzureTtsClient {
     const config = { endpoint, region, subscriptionKey, outputFormat, signal, timeoutMs };
     return synthesizeSpeech(ssml, config);
   }
+
+  async synthesizeSsml(ssml: string): Promise<SsmlSynthesisResult> {
+    const { region, subscriptionKey, outputFormat, signal, timeoutMs } = this.#options;
+    const endpoint = this.#options.endpoint?.trim() || ENDPOINT_TEMPLATE.replace("{region}", region);
+    this.#options.logger?.debug?.("Using Azure TTS endpoint:", endpoint);
+
+    return synthesizeSsml(ssml, { endpoint, region, subscriptionKey, outputFormat, signal, timeoutMs });
+  }
 }

package/src/index.ts

@@ -2,7 +2,23 @@
  * azure-tts-client: Azure Text-to-Speech client for SSML playback.
  */
 
-export type { AzureTtsClientOptions, AzureTtsLogger, TtsConfig } from "./types.ts";
+export type {
+  AzureTtsClientOptions,
+  AzureTtsLogger,
+  SsmlSynthesisBookmark,
+  SsmlSynthesisBoundary,
+  SsmlSynthesisResult,
+  SsmlSynthesisViseme,
+  TtsConfig,
+} from "./types.ts";
 export { AzureTtsError, AzureTtsSdkError } from "./errors.ts";
 export { AzureTtsClient } from "./client.ts";
 export { synthesizeSpeech } from "./synthesis.ts";
+export { synthesizeSsml } from "./synthesis.ts";
+export { fetchAzureVoiceCatalog } from "./voiceCatalog.ts";
+export type {
+  AzureVoiceCatalog,
+  AzureVoiceCatalogVoice,
+  FetchedAzureVoiceCatalogMetadata,
+  FetchAzureVoiceCatalogOptions,
+} from "./voiceCatalog.ts";

package/src/synthesis.ts

@@ -1,7 +1,7 @@
 import * as SpeechSDK from "microsoft-cognitiveservices-speech-sdk";
 import { createSpeechSdkError } from "./errors.ts";
 import { createSpeechConfig } from "./speechConfig.ts";
-import type { TtsConfig } from "./types.ts";
+import type { SsmlSynthesisResult, TtsConfig } from "./types.ts";
 
 function closeSpeechResources(speechConfig: SpeechSDK.SpeechConfig, synthesizer: SpeechSDK.SpeechSynthesizer): void {
   try {
@@ -13,7 +13,9 @@ function closeSpeechResources(speechConfig: SpeechSDK.SpeechConfig, synthesizer:
   } catch {}
 }
 
-export async function synthesizeSpeech(ssml: string, config: TtsConfig): Promise<ArrayBuffer> {
+const ticksToMilliseconds = (ticks: number): number => Math.max(0, ticks) / 10_000;
+
+export async function synthesizeSsml(ssml: string, config: TtsConfig): Promise<SsmlSynthesisResult> {
   if (config.signal?.aborted) {
     throw createSpeechSdkError("Speech synthesis was cancelled.");
   }
@@ -21,7 +23,7 @@ export async function synthesizeSpeech(ssml: string, config: TtsConfig): Promise
   const speechConfig = createSpeechConfig(config);
   const synthesizer = new SpeechSDK.SpeechSynthesizer(speechConfig, null);
 
-  return await new Promise<ArrayBuffer>((resolve, reject) => {
+  return await new Promise<SsmlSynthesisResult>((resolve, reject) => {
     let resourcesClosed = false;
     let settled = false;
     let timeout: ReturnType<typeof setTimeout> | undefined;
@@ -43,6 +45,23 @@ export async function synthesizeSpeech(ssml: string, config: TtsConfig): Promise
       reject(createSpeechSdkError(error));
     };
 
+    const boundaries: SsmlSynthesisResult["boundaries"] = [];
+    const visemes: SsmlSynthesisResult["visemes"] = [];
+    const bookmarks: SsmlSynthesisResult["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: SpeechSDK.SpeechSynthesisResult) => {
       if (settled) return;
       const { reason, errorDetails } = result;
@@ -54,7 +73,20 @@ export async function synthesizeSpeech(ssml: string, config: TtsConfig): Promise
       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 {
@@ -74,3 +106,8 @@ export async function synthesizeSpeech(ssml: string, config: TtsConfig): Promise
     }
   });
 }
+
+/** Backward-compatible audio-only synthesis helper. */
+export async function synthesizeSpeech(ssml: string, config: TtsConfig): Promise<ArrayBuffer> {
+  return (await synthesizeSsml(ssml, config)).audioData;
+}

package/src/types.ts

@@ -7,6 +7,35 @@ export interface TtsConfig {
   outputFormat?: string;
 }
 
+export interface SsmlSynthesisBoundary {
+  text: string;
+  audioOffsetMs: number;
+  durationMs: number;
+}
+
+export interface SsmlSynthesisViseme {
+  visemeId: number;
+  audioOffsetMs: number;
+}
+
+export interface SsmlSynthesisBookmark {
+  name: string;
+  audioOffsetMs: number;
+}
+
+/** Audio and Azure Speech synchronization events emitted for one SSML request. */
+export 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[];
+}
+
 export interface AzureTtsLogger {
   debug?: (...args: unknown[]) => void;
   info?: (...args: unknown[]) => void;

package/src/voiceCatalog.ts

@@ -0,0 +1,120 @@
+const AZURE_VOICE_API_VERSION = "2025-10-01";
+
+export interface FetchAzureVoiceCatalogOptions {
+  apiKey: string;
+  region: string | string[];
+}
+
+export interface AzureVoiceCatalogVoice {
+  name: string;
+  locale: string;
+  secondaryLocales?: readonly string[];
+  styles?: readonly string[];
+  regions: readonly string[];
+  status?: "ga" | "preview" | "deprecated";
+}
+
+export interface FetchedAzureVoiceCatalogMetadata {
+  voiceCount: number;
+  generatedAt: string;
+  apiVersion: string;
+  regions: readonly string[];
+}
+
+export interface AzureVoiceCatalog {
+  voices: readonly AzureVoiceCatalogVoice[];
+  metadata: FetchedAzureVoiceCatalogMetadata;
+}
+
+interface AzureVoiceApiRecord {
+  Locale?: unknown;
+  Name?: unknown;
+  SecondaryLocaleList?: unknown;
+  ShortName?: unknown;
+  Status?: unknown;
+  StyleList?: unknown;
+}
+
+function stringValue(value: unknown): string | undefined {
+  return typeof value === "string" && value.trim() ? value.trim() : undefined;
+}
+
+function stringList(value: unknown): string[] {
+  if (!Array.isArray(value)) return [];
+  return [...new Set(value.map(stringValue).filter((item): item is string => item !== undefined))];
+}
+
+function normalizeStatus(value: unknown): AzureVoiceCatalogVoice["status"] {
+  const status = stringValue(value)?.toLowerCase();
+  if (status === "preview" || status === "deprecated" || status === "ga") return status;
+  return undefined;
+}
+
+function normalizeRegions(region: string | string[]): string[] {
+  const regions = Array.isArray(region) ? region : [region];
+  const result = [...new Set(regions.map((item) => item.trim()).filter(Boolean))];
+  if (result.length === 0) throw new TypeError("At least one Azure Speech region is required.");
+  return result;
+}
+
+async function fetchRegionVoices(region: string, apiKey: string): Promise<AzureVoiceApiRecord[]> {
+  const endpoint = `https://${encodeURIComponent(region)}.tts.speech.microsoft.com/cognitiveservices/voices/list`;
+  const response = await fetch(endpoint, {
+    headers: {
+      Accept: "application/json",
+      "Ocp-Apim-Subscription-Key": apiKey,
+    },
+  });
+  if (!response.ok) {
+    throw new Error(`Azure List Voices API request failed for region "${region}" with HTTP ${response.status}.`);
+  }
+  const payload: unknown = await response.json();
+  if (!Array.isArray(payload)) throw new Error(`Azure List Voices API returned an invalid response for "${region}".`);
+  return payload.filter((item): item is AzureVoiceApiRecord => Boolean(item && typeof item === "object"));
+}
+
+/** Fetches and deduplicates the current Azure Speech voice catalog for one or more regions. */
+export async function fetchAzureVoiceCatalog(options: FetchAzureVoiceCatalogOptions): Promise<AzureVoiceCatalog> {
+  if (!options || typeof options.apiKey !== "string" || !options.apiKey.trim())
+    throw new TypeError("An Azure Speech API key is required.");
+  const regions = normalizeRegions(options.region);
+  const payloads = await Promise.all(regions.map((region) => fetchRegionVoices(region, options.apiKey)));
+  const voices = new Map<string, AzureVoiceCatalogVoice>();
+
+  for (let regionIndex = 0; regionIndex < payloads.length; regionIndex += 1) {
+    const region = regions[regionIndex];
+    for (const record of payloads[regionIndex]) {
+      const name = stringValue(record.ShortName) ?? stringValue(record.Name);
+      const locale = stringValue(record.Locale);
+      if (!name || !locale) continue;
+      const key = name.toLowerCase();
+      const existing = voices.get(key);
+      const secondaryLocales = stringList(record.SecondaryLocaleList);
+      const styles = stringList(record.StyleList);
+      const status = normalizeStatus(record.Status);
+      const merged: AzureVoiceCatalogVoice = {
+        name: existing?.name ?? name,
+        locale: existing?.locale ?? locale,
+        regions: [...new Set([...(existing?.regions ?? []), region])],
+      };
+      const mergedSecondaryLocales = [...new Set([...(existing?.secondaryLocales ?? []), ...secondaryLocales])];
+      if (mergedSecondaryLocales.length > 0) merged.secondaryLocales = mergedSecondaryLocales;
+      const mergedStyles = [...new Set([...(existing?.styles ?? []), ...styles])];
+      if (mergedStyles.length > 0) merged.styles = mergedStyles;
+      if (status) merged.status = status;
+      else if (existing?.status) merged.status = existing.status;
+      voices.set(key, merged);
+    }
+  }
+
+  const sortedVoices = [...voices.values()].sort((first, second) => first.name.localeCompare(second.name));
+  return {
+    voices: sortedVoices,
+    metadata: {
+      voiceCount: sortedVoices.length,
+      generatedAt: new Date().toISOString(),
+      apiVersion: AZURE_VOICE_API_VERSION,
+      regions,
+    },
+  };
+}

package/test/synthesis.test.ts

@@ -1,7 +1,7 @@
 import assert from "node:assert/strict";
 import test, { type TestContext } from "node:test";
 import * as SpeechSDK from "microsoft-cognitiveservices-speech-sdk";
-import { synthesizeSpeech } from "../src/index.ts";
+import { synthesizeSpeech, synthesizeSsml } from "../src/index.ts";
 
 const endpoint = "https://speech.example.test/cognitiveservices/v1";
 const subscriptionKey = "subscription-key";
@@ -51,6 +51,45 @@ test("synthesizeSpeech aborts the SDK request and settles its promise", async (t
   assert.equal(closeCount, 1);
 });
 
+test("synthesizeSsml returns word boundaries, visemes, bookmarks, and duration", async (t) => {
+  const originalFromEndpoint = SpeechSDK.SpeechConfig.fromEndpoint;
+  t.mock.method(SpeechSDK.SpeechConfig, "fromEndpoint", (speechEndpoint, key) =>
+    originalFromEndpoint(speechEndpoint, String(key)),
+  );
+  t.mock.method(
+    SpeechSDK.SpeechSynthesizer.prototype,
+    "speakSsmlAsync",
+    function (this: SpeechSDK.SpeechSynthesizer, _ssml, callback) {
+      this.wordBoundary?.(this, {
+        text: "Hello",
+        audioOffset: 1_000_000,
+        duration: 250_000,
+      } as SpeechSDK.SpeechSynthesisWordBoundaryEventArgs);
+      this.visemeReceived?.(this, { visemeId: 4, audioOffset: 2_000_000 } as SpeechSDK.SpeechSynthesisVisemeEventArgs);
+      this.bookmarkReached?.(this, {
+        text: "chapter-1",
+        audioOffset: 3_000_000,
+      } as SpeechSDK.SpeechSynthesisBookmarkEventArgs);
+      callback?.({
+        audioData: new ArrayBuffer(2),
+        audioDuration: 4_000_000,
+        errorDetails: "",
+        reason: SpeechSDK.ResultReason.SynthesizingAudioCompleted,
+      } as SpeechSDK.SpeechSynthesisResult);
+    },
+  );
+  t.mock.method(SpeechSDK.SpeechSynthesizer.prototype, "close", () => undefined);
+
+  const result = await synthesizeSsml("<speak>Hello</speak>", { endpoint, subscriptionKey, region });
+
+  assert.deepEqual(result.boundaries, [{ text: "Hello", audioOffsetMs: 100, durationMs: 25 }]);
+  assert.deepEqual(result.wordBoundary, result.boundaries);
+  assert.deepEqual(result.wordBoundaries, result.boundaries);
+  assert.deepEqual(result.visemes, [{ visemeId: 4, audioOffsetMs: 200 }]);
+  assert.deepEqual(result.bookmarks, [{ name: "chapter-1", audioOffsetMs: 300 }]);
+  assert.equal(result.durationMs, 400);
+});
+
 test("synthesizeSpeech aborts the SDK request on timeout and settles its promise", async (t) => {
   let closeCount = 0;
   let resultCallback: ((result: SpeechSDK.SpeechSynthesisResult) => void) | undefined;

package/test/voice-catalog.test.ts

@@ -0,0 +1,51 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+import { fetchAzureVoiceCatalog } from "../src/index.ts";
+
+test("fetchAzureVoiceCatalog fetches, merges, and annotates voices across regions", async () => {
+  const originalFetch = globalThis.fetch;
+  const requests: string[] = [];
+  globalThis.fetch = async (input) => {
+    const url = String(input);
+    requests.push(url);
+    const region = url.split(".")[0]?.replace("https://", "");
+    return new Response(
+      JSON.stringify(
+        region === "eastus"
+          ? [
+              { Locale: "fil-PH", ShortName: "fil-PH-BlessicaNeural", StyleList: ["cheerful"] },
+              { Locale: "en-US", Name: "en-US-JennyNeural", StyleList: ["chat"] },
+            ]
+          : [{ Locale: "fil-PH", ShortName: "fil-PH-BlessicaNeural", StyleList: ["sad"] }],
+      ),
+      { status: 200, headers: { "content-type": "application/json" } },
+    );
+  };
+  try {
+    const catalog = await fetchAzureVoiceCatalog({ apiKey: "secret", region: ["eastus", "japaneast"] });
+    assert.equal(requests.length, 2);
+    assert.deepEqual(catalog.metadata.regions, ["eastus", "japaneast"]);
+    assert.equal(catalog.metadata.voiceCount, 2);
+    assert.deepEqual(
+      catalog.voices.find(({ name }) => name === "fil-PH-BlessicaNeural"),
+      {
+        name: "fil-PH-BlessicaNeural",
+        locale: "fil-PH",
+        styles: ["cheerful", "sad"],
+        regions: ["eastus", "japaneast"],
+      },
+    );
+  } finally {
+    globalThis.fetch = originalFetch;
+  }
+});
+
+test("fetchAzureVoiceCatalog reports HTTP and payload failures", async () => {
+  const originalFetch = globalThis.fetch;
+  globalThis.fetch = async () => new Response("no", { status: 401 });
+  try {
+    await assert.rejects(fetchAzureVoiceCatalog({ apiKey: "secret", region: "eastus" }), /HTTP 401/);
+  } finally {
+    globalThis.fetch = originalFetch;
+  }
+});