ssml-builder-js 2.10.0 → 2.12.0

package/dist/core.mjs

@@ -11,10 +11,11 @@ import {
   mapSsmlTextNodes,
   normalizeAzureLanguage,
   parseSsml,
+  splitSsmlDocument,
   validateAzureSsml,
   validateSsml,
   validateSsmlStructureIntegrity
-} from "./chunk-GW6CKHXF.mjs";
+} from "./chunk-4BVNAUVR.mjs";
 import "./chunk-6S5ODO6A.mjs";
 export {
   areAzureLanguagesEquivalent,
@@ -29,6 +30,7 @@ export {
   mapSsmlTextNodes,
   normalizeAzureLanguage,
   parseSsml,
+  splitSsmlDocument,
   validateAzureSsml,
   validateSsml,
   validateSsmlStructureIntegrity

package/dist/elements.js

@@ -956,6 +956,9 @@ var AZURE_VOICE_DEFINITIONS = [
   },
   { name: "es-ES-ElviraNeural", locale: "es-ES" },
   { name: "fil-PH-AngeloNeural", locale: "fil-PH" },
+  { name: "fil-PH-Angelo:DragonHDLatestNeural", locale: "fil-PH" },
+  { name: "fil-PH-BlessicaNeural", locale: "fil-PH" },
+  { name: "fil-PH-Blessica:DragonHDLatestNeural", locale: "fil-PH" },
   { name: "fr-FR-DeniseNeural", locale: "fr-FR", styles: ["cheerful", "sad"] },
   { name: "fr-FR-HenriNeural", locale: "fr-FR", styles: ["cheerful", "sad"] },
   { name: "id-ID-GadisNeural", locale: "id-ID" },
@@ -1050,6 +1053,18 @@ var ALLOWED_SILENCE_TYPES = /* @__PURE__ */ new Set([
   "Enumerationcomma"
 ]);
 var ALLOWED_VISEME_TYPES = /* @__PURE__ */ new Set(["redlips_front", "FacialExpression"]);
+var DEFAULT_PREVIEW_TAGS = /* @__PURE__ */ new Set(["mstts:voiceconversion"]);
+function featureStatusForTag(name, options) {
+  const tagName = canonicalTagName(name);
+  const configured = Object.entries(options.tagStatuses ?? {}).find(
+    ([candidate]) => canonicalTagName(candidate) === tagName
+  )?.[1];
+  if (configured) return configured;
+  if ((options.previewTags ?? [...DEFAULT_PREVIEW_TAGS]).some((candidate) => canonicalTagName(candidate) === tagName))
+    return "preview";
+  if ((options.deprecatedTags ?? []).some((candidate) => canonicalTagName(candidate) === tagName)) return "deprecated";
+  return void 0;
+}
 function decodeAttribute(value) {
   return value.replace(
     /&(?:amp|apos|gt|lt|quot);/gi,
@@ -1118,6 +1133,7 @@ function tokenizeElements(source) {
       attributes,
       childElementIndex,
       end,
+      depth: openElements.length + 1,
       name: tokenName,
       parentName: parent?.name,
       parentVoiceName,
@@ -1166,9 +1182,9 @@ function isValidAzureAudioDuration(value) {
   return Number(clock[1]) > 0 || Number(clock[2]) > 0 || Number(clock[3]) > 0 || Number(clock[4] ?? 0) > 0;
 }
 function isValidAzureBackgroundAudioDuration(value) {
-  const match = /^(\d+(?:\.\d+)?)(ms|s)?$/i.exec(value.trim());
+  const match = /^(\d+)$/.exec(value.trim());
   if (!match) return false;
-  const milliseconds = Number(match[1]) * (match[2]?.toLowerCase() === "s" ? 1e3 : 1);
+  const milliseconds = Number(match[1]);
   return Number.isFinite(milliseconds) && milliseconds >= 0 && milliseconds <= 1e4;
 }
 function attr(token, name) {
@@ -1298,13 +1314,18 @@ function validateAudioSource(token, source, diagnostics, options, elementName2)
     addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> must be an absolute HTTP(S) URL.`);
     return;
   }
+  if (parsed.username || parsed.password)
+    addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> must not contain URL credentials.`);
   if (parsed.protocol !== "https:" && !(options.allowHttpAudio && parsed.protocol === "http:"))
     addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> must use HTTPS.`);
   const isAllowedOrigin = options.allowedAudioOrigins?.some((allowedOrigin) => {
     try {
-      return new URL(allowedOrigin).origin === parsed.origin;
+      const configured = new URL(allowedOrigin);
+      if (configured.protocol !== "https:" && configured.protocol !== "http:" || configured.username || configured.password || configured.pathname !== "/" || configured.search || configured.hash)
+        return false;
+      return configured.origin === parsed.origin;
     } catch {
-      return allowedOrigin === parsed.origin;
+      return false;
     }
   }) ?? false;
   if (options.allowedAudioOrigins && !isAllowedOrigin)
@@ -1319,6 +1340,25 @@ function validateAudioSource(token, source, diagnostics, options, elementName2)
 }
 function validateElement(token, source, diagnostics, voiceName, options, voiceCatalog) {
   const name = token.name.toLowerCase();
+  const tagStatus = featureStatusForTag(token.name, options);
+  if (tagStatus === "preview")
+    addDiagnostic(
+      diagnostics,
+      source,
+      token.start,
+      `<${token.name}> is an Azure Speech preview feature and may change or require preview access.`,
+      "warning",
+      "azure-preview-tag"
+    );
+  if (tagStatus === "deprecated")
+    addDiagnostic(
+      diagnostics,
+      source,
+      token.start,
+      `<${token.name}> is deprecated by Azure Speech; migrate to a supported alternative.`,
+      "info",
+      "azure-deprecated-tag"
+    );
   if (name === "voice" && !attr(token, "name")?.trim())
     addDiagnostic(diagnostics, source, token.start, '<voice> requires a non-empty "name" attribute.');
   if (name === "break") {
@@ -1493,6 +1533,9 @@ function validateAzureSsml(ssml, options = {}) {
   const maxLength = options.maxLength ?? 1e4;
   if (ssml.length > maxLength)
     addDiagnostic(diagnostics, ssml, maxLength, `SSML exceeds the maximum length of ${maxLength} characters.`);
+  if (options.maxXmlDepth !== void 0 && (!Number.isInteger(options.maxXmlDepth) || options.maxXmlDepth <= 0)) {
+    addDiagnostic(diagnostics, ssml, 0, "maxXmlDepth must be a positive integer.");
+  }
   try {
     parseSsml(ssml);
   } catch (error) {
@@ -1502,6 +1545,18 @@ function validateAzureSsml(ssml, options = {}) {
     return diagnostics;
   }
   const tokens = tokenizeElements(ssml);
+  if (options.maxXmlDepth !== void 0) {
+    for (const token of tokens) {
+      if (token.depth > options.maxXmlDepth) {
+        addDiagnostic(
+          diagnostics,
+          ssml,
+          token.start,
+          `XML nesting depth ${token.depth} exceeds the configured maximum of ${options.maxXmlDepth}.`
+        );
+      }
+    }
+  }
   const speak = tokens.find((token) => token.name.toLowerCase() === "speak");
   const voices = tokens.filter((token) => token.name.toLowerCase() === "voice");
   const backgroundAudioTokens = tokens.filter((token) => token.name.toLowerCase() === "mstts:backgroundaudio");
@@ -1530,6 +1585,24 @@ function validateAzureSsml(ssml, options = {}) {
     const name = attr(token, "name")?.trim();
     const language = attr(token, "xml:lang")?.trim() || (speak ? attr(speak, "xml:lang")?.trim() : void 0);
     const definition = name ? voiceCatalog.get(name.toLowerCase()) : void 0;
+    if (name && definition?.status === "preview")
+      addDiagnostic(
+        diagnostics,
+        ssml,
+        token.start,
+        `Voice "${name}" is an Azure Speech preview voice and may change or require preview access.`,
+        "warning",
+        "azure-preview-voice"
+      );
+    if (name && definition?.status === "deprecated")
+      addDiagnostic(
+        diagnostics,
+        ssml,
+        token.start,
+        `Voice "${name}" is deprecated by Azure Speech; migrate to a supported voice.`,
+        "info",
+        "azure-deprecated-voice"
+      );
     if (name && !definition && policySeverity)
       addDiagnostic(
         diagnostics,
@@ -2295,6 +2368,7 @@ var ssmlPresets_exports = {};
 __export(ssmlPresets_exports, {
   AUDIO_DURATION_DESCRIPTIONS: () => AUDIO_DURATION_DESCRIPTIONS,
   AUDIO_DURATION_PRESETS: () => AUDIO_DURATION_PRESETS,
+  BACKGROUND_AUDIO_DURATION_PRESETS: () => BACKGROUND_AUDIO_DURATION_PRESETS,
   BREAK_STRENGTH_PRESETS: () => BREAK_STRENGTH_PRESETS,
   BREAK_TIME_DESCRIPTIONS: () => BREAK_TIME_DESCRIPTIONS,
   BREAK_TIME_PRESETS: () => BREAK_TIME_PRESETS,
@@ -2376,6 +2450,9 @@ var AZURE_VOICE_STYLE_MAP = {
   ],
   "es-ES-ElviraNeural": [],
   "fil-PH-AngeloNeural": [],
+  "fil-PH-Angelo:DragonHDLatestNeural": [],
+  "fil-PH-BlessicaNeural": [],
+  "fil-PH-Blessica:DragonHDLatestNeural": [],
   "fr-FR-DeniseNeural": ["cheerful", "sad"],
   "fr-FR-HenriNeural": ["cheerful", "sad"],
   "id-ID-GadisNeural": [],
@@ -2653,6 +2730,7 @@ var SAY_AS_PRESETS = [
 var LANGUAGE_PRESETS = ["ja-JP", "en-US", "de-DE", "fr-FR"];
 var SILENCE_VALUE_PRESETS = ["300ms", "500ms", "1s"];
 var AUDIO_DURATION_PRESETS = ["5s", "10s", "30s"];
+var BACKGROUND_AUDIO_DURATION_PRESETS = ["0", "500", "1000", "3000", "10000"];
 var SILENCE_TYPE_PRESETS = [
   "Leading",
   "Tailing",
@@ -2714,8 +2792,8 @@ var SSML_ATTRIBUTE_PRESETS = {
   },
   "mstts:backgroundaudio": {
     volume: PROSODY_VOLUME_PRESETS,
-    fadein: AUDIO_DURATION_PRESETS,
-    fadeout: AUDIO_DURATION_PRESETS
+    fadein: BACKGROUND_AUDIO_DURATION_PRESETS,
+    fadeout: BACKGROUND_AUDIO_DURATION_PRESETS
   },
   silence: {
     type: SILENCE_TYPE_PRESETS,
@@ -3220,7 +3298,7 @@ var SSML_COMPLETION_SNIPPETS = [
   },
   {
     label: "mstts:backgroundaudio",
-    insertText: `<mstts:backgroundaudio src="\${1:https://example.com/audio.mp3}" volume="\${2:-3dB}" />`
+    insertText: `<mstts:backgroundaudio src="\${1:https://example.com/audio.mp3}" volume="\${2:70}" fadein="\${3:1000}" fadeout="\${4:1000}" />`
   },
   {
     label: "mstts:ttsembedding",
@@ -3999,18 +4077,18 @@ var SSML_TAG_DEFINITIONS = [
       },
       {
         name: "volume",
-        description: "The background audio volume, for example `-3dB` or `medium`.",
-        example: "-3dB"
+        description: "The background audio volume from 0 to 100.",
+        example: "70"
       },
       {
         name: "fadein",
-        description: "The fade-in duration, for example `1s`.",
-        example: "1s"
+        description: "The fade-in duration in milliseconds from 0 to 10000.",
+        example: "1000"
       },
       {
         name: "fadeout",
-        description: "The fade-out duration, for example `500ms`.",
-        example: "500ms"
+        description: "The fade-out duration in milliseconds from 0 to 10000.",
+        example: "1000"
       }
     ]
   },