ssml-builder-js 2.10.0 → 2.12.0

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,
@@ -2668,6 +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" },
@@ -2762,6 +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,
@@ -2830,6 +2850,7 @@ function tokenizeElements(source) {
       attributes,
       childElementIndex,
       end,
+      depth: openElements.length + 1,
       name: tokenName,
       parentName: parent?.name,
       parentVoiceName,
@@ -2878,9 +2899,9 @@ 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) {
@@ -3010,13 +3031,18 @@ 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, `<${elementName2} src> must use HTTPS.`);
   const isAllowedOrigin = options.allowedAudioOrigins?.some((allowedOrigin) => {
     try {
-      return new URL(allowedOrigin).origin === parsed.origin;
+      const configured = new URL(allowedOrigin);
+      if (configured.protocol !== "https:" && configured.protocol !== "http:" || configured.username || configured.password || configured.pathname !== "/" || configured.search || configured.hash)
+        return false;
+      return configured.origin === parsed.origin;
     } catch {
-      return allowedOrigin === parsed.origin;
+      return false;
     }
   }) ?? false;
   if (options.allowedAudioOrigins && !isAllowedOrigin)
@@ -3031,6 +3057,25 @@ function validateAudioSource(token, source, diagnostics, options, elementName2)
 }
 function validateElement(token, source, diagnostics, voiceName, options, voiceCatalog) {
   const name = token.name.toLowerCase();
+  const tagStatus = featureStatusForTag(token.name, options);
+  if (tagStatus === "preview")
+    addDiagnostic(
+      diagnostics,
+      source,
+      token.start,
+      `<${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") {
@@ -3205,6 +3250,9 @@ 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) {
@@ -3214,6 +3262,18 @@ 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 ${options.maxXmlDepth}.`
+        );
+      }
+    }
+  }
   const speak = tokens.find((token) => token.name.toLowerCase() === "speak");
   const voices = tokens.filter((token) => token.name.toLowerCase() === "voice");
   const backgroundAudioTokens = tokens.filter((token) => token.name.toLowerCase() === "mstts:backgroundaudio");
@@ -3242,6 +3302,24 @@ function validateAzureSsml(ssml, options = {}) {
     const name = attr(token, "name")?.trim();
     const language = attr(token, "xml:lang")?.trim() || (speak ? attr(speak, "xml:lang")?.trim() : void 0);
     const definition = name ? voiceCatalog.get(name.toLowerCase()) : void 0;
+    if (name && definition?.status === "preview")
+      addDiagnostic(
+        diagnostics,
+        ssml,
+        token.start,
+        `Voice "${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,
@@ -4929,7 +5007,7 @@ 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",
@@ -5630,18 +5708,18 @@ 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"
       }
     ]
   },
@@ -6295,6 +6373,175 @@ 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;
+  });
+}
+var VISUAL_ELEMENT_FIELDS = {
+  voice: [
+    { key: "name", label: "Voice name" },
+    { key: "effect", label: "Effect" }
+  ],
+  prosody: [
+    { key: "rate", label: "Rate" },
+    { key: "pitch", label: "Pitch" },
+    { key: "volume", label: "Volume" },
+    { key: "contour", label: "Contour" },
+    { key: "range", label: "Range" }
+  ],
+  "express-as": [
+    { key: "style", label: "Style" },
+    { key: "styleDegree", label: "Style degree" },
+    { key: "role", label: "Role" }
+  ],
+  expressAs: [
+    { key: "style", label: "Style" },
+    { key: "styleDegree", label: "Style degree" },
+    { key: "role", label: "Role" }
+  ],
+  "mstts:express-as": [
+    { key: "style", label: "Style" },
+    { key: "styleDegree", label: "Style degree" },
+    { key: "role", label: "Role" }
+  ],
+  "say-as": [
+    { key: "interpretAs", label: "Interpret as" },
+    { key: "format", label: "Format" },
+    { key: "detail", label: "Detail" }
+  ],
+  sayAs: [
+    { key: "interpretAs", label: "Interpret as" },
+    { key: "format", label: "Format" },
+    { key: "detail", label: "Detail" }
+  ],
+  phoneme: [
+    { key: "alphabet", label: "Alphabet" },
+    { key: "ph", label: "Pronunciation" }
+  ],
+  emphasis: [{ key: "level", label: "Level" }],
+  audio: [
+    { key: "src", label: "Source URL" },
+    { key: "desc", label: "Description" },
+    { key: "clipBegin", label: "Clip begin" },
+    { key: "clipEnd", label: "Clip end" },
+    { key: "speed", label: "Speed" },
+    { key: "repeatCount", label: "Repeat count" },
+    { key: "repeatDuration", label: "Repeat duration" },
+    { key: "soundLevel", label: "Sound level" }
+  ],
+  mark: [{ key: "name", label: "Mark name" }],
+  bookmark: [{ key: "mark", label: "Bookmark name" }],
+  sub: [{ key: "alias", label: "Alias" }],
+  lang: [{ key: "lang", label: "Language" }],
+  lexicon: [{ key: "uri", label: "Lexicon URI" }],
+  "mstts:silence": [
+    { key: "typeValue", label: "Silence type" },
+    { key: "value", label: "Value" }
+  ],
+  silence: [
+    { key: "typeValue", label: "Silence type" },
+    { key: "value", label: "Value" }
+  ],
+  "mstts:audioduration": [{ key: "value", label: "Duration" }],
+  "mstts:ttsembedding": [{ key: "speakerProfileId", label: "Speaker profile ID" }],
+  "mstts:embedding": [
+    { key: "id", label: "Embedding ID" },
+    { key: "speakerProfileId", label: "Speaker profile ID" }
+  ],
+  "mstts:voiceconversion": [
+    { key: "url", label: "Source URL" },
+    { key: "profile", label: "Profile" },
+    { key: "speakerProfileId", label: "Speaker profile ID" }
+  ]
+};
+function getElementFields(element) {
+  return VISUAL_ELEMENT_FIELDS[elementLabel(element)] ?? [];
+}
+function VisualElementInspector({
+  document: document2,
+  element,
+  path,
+  readOnly,
+  commit
+}) {
+  const fields = getElementFields(element);
+  return /* @__PURE__ */ React.createElement("fieldset", null, /* @__PURE__ */ React.createElement("legend", null, `<${elementLabel(element)}>`), fields.length === 0 ? /* @__PURE__ */ React.createElement("p", null, "This element is preserved in the visual tree. Edit its attributes in Code mode.") : fields.map((field) => {
+    const value = element[field.key];
+    const inputValue = value === void 0 ? "" : String(value);
+    const inputId = `ssml-visual-${field.key}`;
+    return /* @__PURE__ */ React.createElement("label", { key: field.key, htmlFor: inputId }, field.label, field.multiline ? /* @__PURE__ */ React.createElement(
+      "textarea",
+      {
+        id: inputId,
+        value: inputValue,
+        readOnly,
+        onChange: (event) => commit(updateOptionalElementProperty(document2, path, field.key, event.target.value))
+      }
+    ) : /* @__PURE__ */ React.createElement(
+      "input",
+      {
+        id: inputId,
+        value: inputValue,
+        readOnly,
+        onChange: (event) => commit(updateOptionalElementProperty(document2, path, field.key, event.target.value))
+      }
+    ));
+  }));
+}
+function getElementAncestors(nodes, path, ancestors = []) {
+  const [index, ...rest] = path;
+  const node = nodes[index];
+  if (!node || !isElement(node)) return ancestors;
+  if (rest.length === 0) return ancestors;
+  return getElementAncestors(node.children ?? [], rest, [...ancestors, node]);
+}
+function buildVisualPreview(document2, leaf, selection) {
+  const start = Math.max(0, Math.min(selection.start, leaf.value.length));
+  const end = Math.max(start, Math.min(selection.end, leaf.value.length));
+  const value = start === end ? leaf.value : leaf.value.slice(start, end);
+  const ancestors = getElementAncestors(document2.children ?? [], leaf.path);
+  const children = ancestors.reduceRight(
+    (current, ancestor) => [{ ...ancestor, children: current }],
+    [value]
+  );
+  return buildSsml({ ...document2, children });
+}
+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,
@@ -6346,6 +6593,7 @@ 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) => {
@@ -6363,7 +6611,79 @@ 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))
+    }
+  ))) : selectedElement && isElement(selectedElement) ? /* @__PURE__ */ React.createElement(
+    VisualElementInspector,
+    {
+      document: document2,
+      element: selectedElement,
+      path: selectedPath ?? [],
+      readOnly,
+      commit
+    }
+  ) : selectedLeaf ? /* @__PURE__ */ React.createElement(React.Fragment, null, /* @__PURE__ */ React.createElement("label", null, "Text", /* @__PURE__ */ React.createElement(
     "textarea",
     {
       value: selectedLeaf.value,
@@ -6391,7 +6711,7 @@ function VisualSsmlEditor({
     "button",
     {
       type: "button",
-      onClick: () => onPreviewSelection(buildSsml({ ...document2, children: [selectedLeaf.value] }))
+      onClick: () => onPreviewSelection(buildVisualPreview(document2, selectedLeaf, selection))
     },
     "Preview selection"
   ))) : /* @__PURE__ */ React.createElement("p", null, "Select an element or text node to edit it."))));