ssml-builder-js 2.9.0 → 2.11.0

package/dist/react.js

@@ -57,6 +57,7 @@ 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,
@@ -138,6 +139,9 @@ 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": [],
@@ -415,6 +419,7 @@ 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",
@@ -476,8 +481,8 @@ 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,
@@ -1724,6 +1729,7 @@ var SSML_ATTRS = {
   MSTTS_XMLNS: "xmlns:mstts",
   NAME: "name",
   VOICE: "voice",
+  SPEAKER: "speaker",
   EFFECT: "effect",
   RATE: "rate",
   PITCH: "pitch",
@@ -1754,6 +1760,11 @@ var SSML_ATTRS = {
   ALIAS: "alias",
   MARK: "mark",
   URI: "uri",
+  ID: "id",
+  MODEL: "model",
+  PROFILE: "profile",
+  URL: "url",
+  SPEAKER_PROFILE_ID: "speakerProfileId",
   TYPE: "type",
   VALUE: "value",
   FADE_IN: "fadein",
@@ -1849,6 +1860,7 @@ function getAttributes(element) {
       break;
     case SSML_TAGS.MSTTS_TURN:
       addAttribute(attributes, SSML_ATTRS.VOICE, element.voice);
+      addAttribute(attributes, SSML_ATTRS.SPEAKER, element.speaker);
       break;
     case SSML_TAGS.MSTTS_BACKGROUND_AUDIO:
       addAttribute(attributes, SSML_ATTRS.SRC, element.src);
@@ -1857,9 +1869,18 @@ function getAttributes(element) {
       addAttribute(attributes, SSML_ATTRS.FADE_OUT, element.fadeOut ?? element.fadeout);
       break;
     case SSML_TAGS.MSTTS_DIALOG:
+      break;
     case SSML_TAGS.MSTTS_TTS_EMBEDDING:
+      addAttribute(attributes, SSML_ATTRS.SPEAKER_PROFILE_ID, element.speakerProfileId);
+      break;
     case SSML_TAGS.MSTTS_EMBEDDING:
+      addAttribute(attributes, SSML_ATTRS.ID, element.id);
+      addAttribute(attributes, SSML_ATTRS.SPEAKER_PROFILE_ID, element.speakerProfileId);
+      break;
     case SSML_TAGS.MSTTS_VOICE_CONVERSION:
+      addAttribute(attributes, SSML_ATTRS.URL, element.url);
+      addAttribute(attributes, SSML_ATTRS.PROFILE, element.profile);
+      addAttribute(attributes, SSML_ATTRS.SPEAKER_PROFILE_ID, element.speakerProfileId);
       break;
     case SSML_TAGS.PARAGRAPH:
     case SSML_TAGS.SENTENCE:
@@ -2432,7 +2453,9 @@ function convertElement(node) {
     case SSML_TAGS.MSTTS_TURN: {
       const element = { type: SSML_TAGS.MSTTS_TURN };
       const voice = readAttribute(attributes, SSML_ATTRS.VOICE);
+      const speaker = readAttribute(attributes, SSML_ATTRS.SPEAKER);
       if (voice !== void 0) element.voice = voice;
+      if (speaker !== void 0) element.speaker = speaker;
       return finishElement(element, node, attributes);
     }
     case SSML_TAGS.MSTTS_BACKGROUND_AUDIO: {
@@ -2449,14 +2472,26 @@ function convertElement(node) {
     }
     case SSML_TAGS.MSTTS_TTS_EMBEDDING: {
       const element = { type: SSML_TAGS.MSTTS_TTS_EMBEDDING };
+      const speakerProfileId = readAttribute(attributes, SSML_ATTRS.SPEAKER_PROFILE_ID);
+      if (speakerProfileId !== void 0) element.speakerProfileId = speakerProfileId;
       return finishElement(element, node, attributes);
     }
     case SSML_TAGS.MSTTS_EMBEDDING: {
       const element = { type: SSML_TAGS.MSTTS_EMBEDDING };
+      const id = readAttribute(attributes, SSML_ATTRS.ID);
+      const speakerProfileId = readAttribute(attributes, SSML_ATTRS.SPEAKER_PROFILE_ID);
+      if (id !== void 0) element.id = id;
+      if (speakerProfileId !== void 0) element.speakerProfileId = speakerProfileId;
       return finishElement(element, node, attributes);
     }
     case SSML_TAGS.MSTTS_VOICE_CONVERSION: {
       const element = { type: SSML_TAGS.MSTTS_VOICE_CONVERSION };
+      const url = readAttribute(attributes, SSML_ATTRS.URL);
+      const profile = readAttribute(attributes, SSML_ATTRS.PROFILE);
+      const speakerProfileId = readAttribute(attributes, SSML_ATTRS.SPEAKER_PROFILE_ID);
+      if (url !== void 0) element.url = url;
+      if (profile !== void 0) element.profile = profile;
+      if (speakerProfileId !== void 0) element.speakerProfileId = speakerProfileId;
       return finishElement(element, node, attributes);
     }
     default: {
@@ -2638,6 +2673,9 @@ 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" },
@@ -2732,6 +2770,18 @@ 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,
@@ -2791,11 +2841,14 @@ function tokenizeElements(source) {
     }
     const selfClosing = /\/\s*>$/.test(raw);
     const parent = openElements[openElements.length - 1];
+    const childElementIndex = parent?.childElementCount;
+    if (parent) parent.childElementCount += 1;
     const parentVoiceName = [...openElements].reverse().find((element) => element.voiceName)?.voiceName;
     const tokenName = nameMatch[1];
     const tokenVoiceName = tokenName.toLowerCase() === "voice" ? attributes.get("name") : tokenName.toLowerCase() === "mstts:turn" ? attributes.get("voice") ?? parentVoiceName : parentVoiceName;
     tokens.push({
       attributes,
+      childElementIndex,
       end,
       name: tokenName,
       parentName: parent?.name,
@@ -2805,6 +2858,7 @@ function tokenizeElements(source) {
     });
     if (!selfClosing) {
       openElements.push({
+        childElementCount: 0,
         name: tokenName,
         voiceName: tokenVoiceName
       });
@@ -2843,6 +2897,12 @@ function isValidAzureAudioDuration(value) {
   if (!clock) return false;
   return Number(clock[1]) > 0 || Number(clock[2]) > 0 || Number(clock[3]) > 0 || Number(clock[4] ?? 0) > 0;
 }
+function isValidAzureBackgroundAudioDuration(value) {
+  const match = /^(\d+)$/.exec(value.trim());
+  if (!match) return false;
+  const milliseconds = Number(match[1]);
+  return Number.isFinite(milliseconds) && milliseconds >= 0 && milliseconds <= 1e4;
+}
 function attr(token, name) {
   return token.attributes.get(name.toLowerCase());
 }
@@ -2972,9 +3032,16 @@ function validateAudioSource(token, source, diagnostics, options, elementName2)
   }
   if (parsed.protocol !== "https:" && !(options.allowHttpAudio && parsed.protocol === "http:"))
     addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> must use HTTPS.`);
-  if (options.allowedAudioOrigins && !options.allowedAudioOrigins.includes(parsed.origin))
+  const isAllowedOrigin = options.allowedAudioOrigins?.some((allowedOrigin) => {
+    try {
+      return new URL(allowedOrigin).origin === parsed.origin;
+    } catch {
+      return allowedOrigin === parsed.origin;
+    }
+  }) ?? false;
+  if (options.allowedAudioOrigins && !isAllowedOrigin)
     addDiagnostic(diagnostics, source, token.start, `<${elementName2} src> origin "${parsed.origin}" is not allowed.`);
-  else if (!options.allowExternalAudio)
+  else if (!isAllowedOrigin && !options.allowExternalAudio)
     addDiagnostic(
       diagnostics,
       source,
@@ -2984,6 +3051,25 @@ function validateAudioSource(token, source, diagnostics, options, elementName2)
 }
 function validateElement(token, source, diagnostics, voiceName, options, voiceCatalog) {
   const name = token.name.toLowerCase();
+  const tagStatus = featureStatusForTag(token.name, options);
+  if (tagStatus === "preview")
+    addDiagnostic(
+      diagnostics,
+      source,
+      token.start,
+      `<${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,
+      `<${token.name}> is deprecated by Azure Speech; migrate to a supported alternative.`,
+      "info",
+      "azure-deprecated-tag"
+    );
   if (name === "voice" && !attr(token, "name")?.trim())
     addDiagnostic(diagnostics, source, token.start, '<voice> requires a non-empty "name" attribute.');
   if (name === "break") {
@@ -3104,28 +3190,40 @@ function validateElement(token, source, diagnostics, voiceName, options, voiceCa
     validateAudioSource(token, source, diagnostics, options, "audio");
   }
   if (name === "mstts:turn") {
-    if (!attr(token, "voice")?.trim())
-      addDiagnostic(diagnostics, source, token.start, '<mstts:turn> requires a non-empty "voice" attribute.');
+    if (!attr(token, "voice")?.trim() && !attr(token, "speaker")?.trim())
+      addDiagnostic(
+        diagnostics,
+        source,
+        token.start,
+        '<mstts:turn> requires a non-empty "voice" or "speaker" attribute.'
+      );
     if (token.parentName?.toLowerCase() !== "mstts:dialog")
       addDiagnostic(diagnostics, source, token.start, "<mstts:turn> is only allowed directly inside <mstts:dialog>.");
   }
   if (name === "mstts:backgroundaudio") {
     validateAudioSource(token, source, diagnostics, options, "mstts:backgroundaudio");
     const volume = attr(token, "volume");
-    if (volume && !/^(silent|x-soft|soft|medium|loud|x-loud|[+-]?\d+(?:\.\d+)?(?:dB|%))$/i.test(volume.trim()))
+    if (volume !== void 0 && (!/^\d+(?:\.\d+)?$/.test(volume.trim()) || Number(volume) > 100))
       addDiagnostic(diagnostics, source, token.start, `Unsupported <mstts:backgroundaudio volume> value "${volume}".`);
     for (const [attribute, value] of [
       ["fadein", attr(token, "fadein")],
       ["fadeout", attr(token, "fadeout")]
     ]) {
-      if (value && !isValidAzureAudioDuration(value))
+      if (value !== void 0 && !isValidAzureBackgroundAudioDuration(value))
         addDiagnostic(
           diagnostics,
           source,
           token.start,
-          `<mstts:backgroundaudio ${attribute}> must be a positive duration such as "500ms" or "10s".`
+          `<mstts:backgroundaudio ${attribute}> must be between 0 and 10000 milliseconds, for example "500ms" or "10s".`
         );
     }
+    if (token.parentName?.toLowerCase() !== "speak" || token.childElementIndex !== 0)
+      addDiagnostic(
+        diagnostics,
+        source,
+        token.start,
+        "<mstts:backgroundaudio> must be the first element directly under <speak>."
+      );
     if (!token.selfClosing)
       addDiagnostic(diagnostics, source, token.start, "<mstts:backgroundaudio> must be self-closing.");
   }
@@ -3157,6 +3255,16 @@ function validateAzureSsml(ssml, options = {}) {
   const tokens = tokenizeElements(ssml);
   const speak = tokens.find((token) => token.name.toLowerCase() === "speak");
   const voices = tokens.filter((token) => token.name.toLowerCase() === "voice");
+  const backgroundAudioTokens = tokens.filter((token) => token.name.toLowerCase() === "mstts:backgroundaudio");
+  for (const [index, token] of backgroundAudioTokens.entries()) {
+    if (index > 0)
+      addDiagnostic(
+        diagnostics,
+        ssml,
+        token.start,
+        "An SSML document can contain at most one <mstts:backgroundaudio> element."
+      );
+  }
   if (!speak || voices.length === 0)
     addDiagnostic(
       diagnostics,
@@ -3173,6 +3281,24 @@ function validateAzureSsml(ssml, options = {}) {
     const name = attr(token, "name")?.trim();
     const language = attr(token, "xml:lang")?.trim() || (speak ? attr(speak, "xml:lang")?.trim() : void 0);
     const definition = name ? voiceCatalog.get(name.toLowerCase()) : void 0;
+    if (name && definition?.status === "preview")
+      addDiagnostic(
+        diagnostics,
+        ssml,
+        token.start,
+        `Voice "${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,
@@ -4860,7 +4986,7 @@ 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",
@@ -5561,18 +5687,18 @@ 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"
       }
     ]
   },
@@ -6226,6 +6352,46 @@ function updateNodesAtPath(nodes, path, update) {
 function updateTextAtPath(document2, path, value) {
   return { ...document2, children: updateNodesAtPath(document2.children ?? [], path, () => value) };
 }
+function getNodeAtPath(nodes, path) {
+  let current;
+  let currentNodes = nodes;
+  for (const index of path) {
+    current = currentNodes[index];
+    if (current === void 0) return void 0;
+    if (!isElement(current)) {
+      currentNodes = [];
+    } else {
+      currentNodes = current.children ?? [];
+    }
+  }
+  return current;
+}
+function updateElementAtPath(document2, path, update) {
+  return {
+    ...document2,
+    children: updateNodesAtPath(document2.children ?? [], path, (node) => isElement(node) ? update(node) : node)
+  };
+}
+function updateElementText(element, value) {
+  return { ...element, children: [value] };
+}
+function updateOptionalElementProperty(document2, path, property, value) {
+  return updateElementAtPath(document2, path, (element) => {
+    const next = { ...element };
+    if (value.trim()) next[property] = value;
+    else delete next[property];
+    return next;
+  });
+}
+function addDialogTurn(document2, path) {
+  return updateElementAtPath(document2, path, (element) => ({
+    ...element,
+    children: [
+      ...element.children ?? [],
+      { type: "mstts:turn", voice: "en-US-JennyNeural", children: ["New dialog turn"] }
+    ]
+  }));
+}
 function wrapTextAtPath(document2, path, start, end, tag, attributes) {
   return {
     ...document2,
@@ -6277,6 +6443,7 @@ function VisualSsmlEditor({
   const [selection, setSelection] = (0, import_react3.useState)({ start: 0, end: 0 });
   const textLeaves = (0, import_react3.useMemo)(() => collectTextLeaves(document2.children ?? []), [document2]);
   const selectedLeaf = textLeaves.find((leaf) => leaf.path.join(".") === selectedPath?.join(".")) ?? textLeaves[0];
+  const selectedElement = selectedPath ? getNodeAtPath(document2.children ?? [], selectedPath) : void 0;
   const diagnostics = (0, import_react3.useMemo)(() => validateAzureSsml(buildSsml(document2)), [document2]);
   const commit = (nextDocument) => onChange?.(nextDocument);
   const applyWrapper = (tag, attributes) => {
@@ -6294,7 +6461,70 @@ function VisualSsmlEditor({
       selectedPath,
       onSelect: setSelectedPath
     }
-  )))), /* @__PURE__ */ React.createElement("div", { className: "ssml-editor-visual-form" }, selectedLeaf ? /* @__PURE__ */ React.createElement(React.Fragment, null, /* @__PURE__ */ React.createElement("label", null, "Text", /* @__PURE__ */ React.createElement(
+  )))), /* @__PURE__ */ React.createElement("div", { className: "ssml-editor-visual-form" }, selectedElement && isElement(selectedElement) && selectedElement.type === "mstts:dialog" ? /* @__PURE__ */ React.createElement("fieldset", null, /* @__PURE__ */ React.createElement("legend", null, "Dialog"), /* @__PURE__ */ React.createElement("p", null, "Add and edit speaker turns in this dialog."), /* @__PURE__ */ React.createElement(
+    "button",
+    {
+      type: "button",
+      disabled: readOnly,
+      onClick: () => commit(addDialogTurn(document2, selectedPath ?? []))
+    },
+    "Add turn"
+  )) : selectedElement && isElement(selectedElement) && selectedElement.type === "mstts:turn" ? /* @__PURE__ */ React.createElement("fieldset", null, /* @__PURE__ */ React.createElement("legend", null, "Dialog turn"), /* @__PURE__ */ React.createElement("label", null, "Voice", /* @__PURE__ */ React.createElement(
+    "input",
+    {
+      value: selectedElement.voice ?? "",
+      readOnly,
+      onChange: (event) => commit(updateOptionalElementProperty(document2, selectedPath ?? [], "voice", event.target.value))
+    }
+  )), /* @__PURE__ */ React.createElement("label", null, "Speaker", /* @__PURE__ */ React.createElement(
+    "input",
+    {
+      value: selectedElement.speaker ?? "",
+      readOnly,
+      onChange: (event) => commit(updateOptionalElementProperty(document2, selectedPath ?? [], "speaker", event.target.value))
+    }
+  )), /* @__PURE__ */ React.createElement("label", null, "Turn text", /* @__PURE__ */ React.createElement(
+    "textarea",
+    {
+      value: selectedElement.children?.filter((child) => typeof child === "string").join("") ?? "",
+      readOnly,
+      onChange: (event) => commit(
+        updateElementAtPath(
+          document2,
+          selectedPath ?? [],
+          (element) => updateElementText(element, event.target.value)
+        )
+      )
+    }
+  ))) : selectedElement && isElement(selectedElement) && selectedElement.type === "mstts:backgroundaudio" ? /* @__PURE__ */ React.createElement("fieldset", null, /* @__PURE__ */ React.createElement("legend", null, "Background audio"), /* @__PURE__ */ React.createElement("label", null, "Source URL", /* @__PURE__ */ React.createElement(
+    "input",
+    {
+      value: selectedElement.src ?? "",
+      readOnly,
+      onChange: (event) => commit(updateOptionalElementProperty(document2, selectedPath ?? [], "src", event.target.value))
+    }
+  )), /* @__PURE__ */ React.createElement("label", null, "Volume", /* @__PURE__ */ React.createElement(
+    "input",
+    {
+      value: selectedElement.volume === void 0 ? "" : String(selectedElement.volume),
+      readOnly,
+      onChange: (event) => commit(updateOptionalElementProperty(document2, selectedPath ?? [], "volume", event.target.value))
+    }
+  )), /* @__PURE__ */ React.createElement("label", null, "Fade in", /* @__PURE__ */ React.createElement(
+    "input",
+    {
+      value: selectedElement.fadeIn === void 0 ? "" : String(selectedElement.fadeIn),
+      readOnly,
+      onChange: (event) => commit(updateOptionalElementProperty(document2, selectedPath ?? [], "fadeIn", event.target.value))
+    }
+  )), /* @__PURE__ */ React.createElement("label", null, "Fade out", /* @__PURE__ */ React.createElement(
+    "input",
+    {
+      value: selectedElement.fadeOut === void 0 ? "" : String(selectedElement.fadeOut),
+      readOnly,
+      onChange: (event) => commit(updateOptionalElementProperty(document2, selectedPath ?? [], "fadeOut", event.target.value))
+    }
+  ))) : selectedLeaf ? /* @__PURE__ */ React.createElement(React.Fragment, null, /* @__PURE__ */ React.createElement("label", null, "Text", /* @__PURE__ */ React.createElement(
     "textarea",
     {
       value: selectedLeaf.value,