@ssml-builder-js/ssml-editor-react 2.7.0 → 2.8.1

package/CHANGELOG.md

@@ -1,5 +1,20 @@
 # @ssml-builder-js/ssml-editor-react
 
+## 2.8.1
+
+### Patch Changes
+
+- ba09a72: Prepare the v2.8.1 patch release.
+- Updated dependencies [ba09a72]
+  - @ssml-builder-js/ssml-core@2.8.1
+
+## 2.8.0
+
+### Patch Changes
+
+- Updated dependencies [c7c8cb4]
+  - @ssml-builder-js/ssml-core@2.8.0
+
 ## 2.7.0
 
 ### Patch Changes

package/dist/index.d.mts

@@ -3,7 +3,7 @@ import { ReactNode, CSSProperties } from 'react';
 import { SsmlDocument } from '@ssml-builder-js/ssml-core';
 import { OnMount, Monaco } from '@monaco-editor/react';
 
-type SsmlEditorInsertionButton = "break" | "emphasis" | "rate" | "pitch" | "volume" | "emotion" | "say-as" | "phoneme" | "audio" | "sub" | "lang" | "mark" | "bookmark" | "mstts:silence" | "mstts:viseme" | (string & {});
+type SsmlEditorInsertionButton = "break" | "emphasis" | "rate" | "pitch" | "volume" | "emotion" | "say-as" | "phoneme" | "audio" | "sub" | "lang" | "mark" | "bookmark" | "mstts:silence" | "mstts:audioduration" | "mstts:viseme" | (string & {});
 type SsmlEditorButton = "help" | SsmlEditorInsertionButton | "undo" | "redo" | "clearAll" | "format" | "decorations" | (string & {});
 type SsmlEditorButtonVisibility = Readonly<Partial<Record<string, boolean>>>;
 
@@ -274,8 +274,8 @@ declare function updateTagAttribute(text: string, tagRange: SsmlTagRange, attrib
 
 type SsmlCodeLensAction = {
     type: "attribute";
-    insertionId: "rate" | "pitch" | "break";
-    attributeName: "rate" | "pitch" | "time";
+    insertionId: "rate" | "pitch" | "break" | "mstts:audioduration";
+    attributeName: "rate" | "pitch" | "time" | "value";
     tagRange: SsmlTagRange;
 } | {
     type: "unwrap" | "delete";

package/dist/index.d.ts

@@ -3,7 +3,7 @@ import { ReactNode, CSSProperties } from 'react';
 import { SsmlDocument } from '@ssml-builder-js/ssml-core';
 import { OnMount, Monaco } from '@monaco-editor/react';
 
-type SsmlEditorInsertionButton = "break" | "emphasis" | "rate" | "pitch" | "volume" | "emotion" | "say-as" | "phoneme" | "audio" | "sub" | "lang" | "mark" | "bookmark" | "mstts:silence" | "mstts:viseme" | (string & {});
+type SsmlEditorInsertionButton = "break" | "emphasis" | "rate" | "pitch" | "volume" | "emotion" | "say-as" | "phoneme" | "audio" | "sub" | "lang" | "mark" | "bookmark" | "mstts:silence" | "mstts:audioduration" | "mstts:viseme" | (string & {});
 type SsmlEditorButton = "help" | SsmlEditorInsertionButton | "undo" | "redo" | "clearAll" | "format" | "decorations" | (string & {});
 type SsmlEditorButtonVisibility = Readonly<Partial<Record<string, boolean>>>;
 
@@ -274,8 +274,8 @@ declare function updateTagAttribute(text: string, tagRange: SsmlTagRange, attrib
 
 type SsmlCodeLensAction = {
     type: "attribute";
-    insertionId: "rate" | "pitch" | "break";
-    attributeName: "rate" | "pitch" | "time";
+    insertionId: "rate" | "pitch" | "break" | "mstts:audioduration";
+    attributeName: "rate" | "pitch" | "time" | "value";
     tagRange: SsmlTagRange;
 } | {
     type: "unwrap" | "delete";

package/dist/index.js

@@ -54,6 +54,8 @@ function isSsmlEditorButtonVisible(buttonVisibility, button) {
 // src/constants/ssmlPresets.ts
 var ssmlPresets_exports = {};
 __export(ssmlPresets_exports, {
+  AUDIO_DURATION_DESCRIPTIONS: () => AUDIO_DURATION_DESCRIPTIONS,
+  AUDIO_DURATION_PRESETS: () => AUDIO_DURATION_PRESETS,
   BREAK_STRENGTH_PRESETS: () => BREAK_STRENGTH_PRESETS,
   BREAK_TIME_DESCRIPTIONS: () => BREAK_TIME_DESCRIPTIONS,
   BREAK_TIME_PRESETS: () => BREAK_TIME_PRESETS,
@@ -86,6 +88,103 @@ __export(ssmlPresets_exports, {
   getExpressAsStyleCategory: () => getExpressAsStyleCategory,
   resolveExpressAsStyles: () => resolveExpressAsStyles
 });
+
+// src/constants/azureVoiceStyleMap.generated.ts
+var AZURE_VOICE_STYLE_MAP = {
+  "de-DE-ConradNeural": ["cheerful", "sad"],
+  "de-DE-KatjaNeural": ["cheerful", "sad"],
+  "en-US-AndrewNeural": ["empathetic", "relieved"],
+  "en-US-GuyNeural": [
+    "angry",
+    "cheerful",
+    "excited",
+    "friendly",
+    "hopeful",
+    "newscast",
+    "sad",
+    "shouting",
+    "terrified",
+    "unfriendly",
+    "whispering"
+  ],
+  "en-US-JennyMultilingualNeural": [
+    "cheerful",
+    "empathetic",
+    "excited",
+    "friendly",
+    "hopeful",
+    "sad",
+    "shouting",
+    "terrified",
+    "unfriendly",
+    "whispering"
+  ],
+  "en-US-JennyNeural": [
+    "assistant",
+    "chat",
+    "customerservice",
+    "newscast",
+    "cheerful",
+    "empathetic",
+    "excited",
+    "friendly",
+    "hopeful",
+    "sad",
+    "shouting",
+    "terrified",
+    "unfriendly",
+    "whispering"
+  ],
+  "es-ES-ElviraNeural": [],
+  "fil-PH-AngeloNeural": [],
+  "fr-FR-DeniseNeural": ["cheerful", "sad"],
+  "fr-FR-HenriNeural": ["cheerful", "sad"],
+  "id-ID-GadisNeural": [],
+  "it-IT-ElsaNeural": ["cheerful", "sad"],
+  "ja-JP-KeitaNeural": ["chat"],
+  "ja-JP-MayuNeural": ["calm", "cheerful", "sad"],
+  "ja-JP-NanamiNeural": ["chat", "customerservice", "cheerful", "whispering", "sad"],
+  "ko-KR-SunHiNeural": ["cheerful", "sad"],
+  "ms-MY-YasminNeural": [],
+  "pt-BR-FranciscaNeural": ["calm"],
+  "ru-RU-SvetlanaNeural": ["cheerful", "sad", "angry", "disgruntled", "embarrassed", "fearful"],
+  "th-TH-PremwadeeNeural": [],
+  "vi-VN-HoaiMyNeural": [],
+  "zh-CN-XiaoxiaoNeural": [
+    "assistant",
+    "chat",
+    "customerservice",
+    "newscast",
+    "cheerful",
+    "empathetic",
+    "excited",
+    "friendly",
+    "hopeful",
+    "sad",
+    "terrified",
+    "whispering",
+    "poetry-reading",
+    "sports_commentary",
+    "sports_commentary_excited",
+    "story"
+  ],
+  "zh-CN-YunxiNeural": [
+    "narration-relaxed",
+    "embarrassed",
+    "fearful",
+    "sad",
+    "disgruntled",
+    "serious",
+    "angry",
+    "depressed",
+    "chat",
+    "cheerful",
+    "assistant"
+  ],
+  "zh-TW-HsiaoChenNeural": []
+};
+
+// src/constants/ssmlPresets.ts
 var SSML_PRESETS = [
   {
     id: "basic",
@@ -185,7 +284,7 @@ function getExpressAsStyleCategory(style) {
   }
   return "other";
 }
-var VOICE_STYLE_MAP = {
+var LEGACY_VOICE_STYLE_MAP = {
   "ja-JP-MayuNeural": ["calm", "cheerful", "sad"],
   "ja-JP-KeitaNeural": ["chat"],
   "ja-JP-NanamiNeural": ["chat", "customerservice", "cheerful", "whispering", "sad"],
@@ -271,6 +370,7 @@ var VOICE_STYLE_MAP = {
   "de-DE-ConradNeural": ["cheerful", "sad"],
   "ru-RU-SvetlanaNeural": ["cheerful", "sad", "angry", "disgruntled", "embarrassed", "fearful"]
 };
+var VOICE_STYLE_MAP = Object.keys(AZURE_VOICE_STYLE_MAP).length > 0 ? AZURE_VOICE_STYLE_MAP : LEGACY_VOICE_STYLE_MAP;
 var VOICE_STYLE_MAP_BY_NORMALIZED_NAME = new Map(
   Object.entries(VOICE_STYLE_MAP).map(([voiceName, styles]) => [voiceName.toLowerCase(), styles])
 );
@@ -313,6 +413,7 @@ 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 SILENCE_TYPE_PRESETS = [
   "Leading",
   "Tailing",
@@ -321,6 +422,11 @@ var SILENCE_TYPE_PRESETS = [
   "Semicolon",
   "Enumerationcomma"
 ];
+var AUDIO_DURATION_DESCRIPTIONS = {
+  "5s": { ja: "5\u79D2", en: "5 seconds" },
+  "10s": { ja: "10\u79D2", en: "10 seconds" },
+  "30s": { ja: "30\u79D2", en: "30 seconds" }
+};
 var PHONEME_ALPHABET_PRESETS = ["ipa", "sapi", "ups", "x-sampa"];
 var VISEME_TYPE_PRESETS = ["redlips_front", "FacialExpression"];
 var SSML_ATTRIBUTE_PRESETS = {
@@ -364,6 +470,9 @@ var SSML_ATTRIBUTE_PRESETS = {
     type: SILENCE_TYPE_PRESETS,
     value: SILENCE_VALUE_PRESETS
   },
+  "mstts:audioduration": {
+    value: AUDIO_DURATION_PRESETS
+  },
   silence: {
     type: SILENCE_TYPE_PRESETS,
     value: SILENCE_VALUE_PRESETS
@@ -910,6 +1019,13 @@ var SSML_HOVER_COPY = {
         parameters: {
           type: { title: "type", description: "\u30D3\u30BC\u30FC\u30E0\u30A4\u30D9\u30F3\u30C8\u306E\u5F62\u5F0F\u3002" }
         }
+      },
+      "mstts:audioduration": {
+        title: "\u97F3\u58F0\u9577",
+        description: "\u5408\u6210\u97F3\u58F0\u306E\u76EE\u6A19\u6642\u9593\u3092\u6307\u5B9A\u3057\u307E\u3059\u3002",
+        parameters: {
+          value: { title: "value", description: "\u76EE\u6A19\u6642\u9593\u3002\u4F8B: `10s`\u3001`5000ms`\u3001`00:00:10`\u3002" }
+        }
       }
     }
   },
@@ -1067,6 +1183,13 @@ var SSML_HOVER_COPY = {
         parameters: {
           type: { title: "type", description: "The viseme event format." }
         }
+      },
+      "mstts:audioduration": {
+        title: "Audio duration",
+        description: "Sets the target duration of synthesized audio.",
+        parameters: {
+          value: { title: "value", description: "The target duration, such as `10s`, `5000ms`, or `00:00:10`." }
+        }
       }
     }
   }
@@ -1604,6 +1727,7 @@ var INTRINSICALLY_EMPTY_ELEMENTS = /* @__PURE__ */ new Set([
   "lexicon",
   "mark",
   "mstts:silence",
+  "mstts:audioduration",
   "mstts:viseme",
   "silence",
   "viseme"
@@ -2450,7 +2574,7 @@ function findSsmlVoiceContext(source, offset) {
 }
 
 // src/hooks/useSsmlEditorState.ts
-var TIMING_INSERTION_TAGS = /* @__PURE__ */ new Set(["break", "mstts:silence"]);
+var TIMING_INSERTION_TAGS = /* @__PURE__ */ new Set(["break", "mstts:silence", "mstts:audioduration"]);
 var PROSODY_INSERTION_TAGS = /* @__PURE__ */ new Set(["prosody", "mstts:express-as", "voice", "emphasis"]);
 var TEXT_INSERTION_TAGS = /* @__PURE__ */ new Set(["sub", "say-as", "phoneme", "w", "lang"]);
 var EMPTY_SELECTION_OVERLAY = {
@@ -3172,6 +3296,10 @@ var SSML_COMPLETION_SNIPPETS = [
     label: "mstts:express-as",
     insertText: `<mstts:express-as style="cheerful">\${1:text}</mstts:express-as>`
   },
+  {
+    label: "mstts:audioduration",
+    insertText: '<mstts:audioduration value="10s" />'
+  },
   {
     label: "sub",
     insertText: `<sub alias="\${1:\u8AAD\u307F}">\${2:\u6F22\u5B57}</sub>`
@@ -3447,7 +3575,7 @@ function registerSsmlCodeLens(monaco, editor, onOpenPopover) {
           break;
         }
         const tag = source.slice(tagStart, tagEnd + 1);
-        const tagName = tag.match(/^<\s*(prosody|break)\b/i)?.[1]?.toLowerCase();
+        const tagName = tag.match(/^<\s*(prosody|break|mstts:audioduration)\b/i)?.[1]?.toLowerCase();
         index = tagEnd + 1;
         if (!tagName) {
           continue;
@@ -3477,7 +3605,22 @@ function registerSsmlCodeLens(monaco, editor, onOpenPopover) {
               elementRange
             })
           );
-        } else if (/\/\s*>$/.test(tag)) {
+        } else if (tagName === "mstts:audioduration" && /\/\s*>$/.test(tag)) {
+          lenses.push(
+            createLens(
+              model,
+              tagStart,
+              tagEnd + 1,
+              `\u26A1 Duration: ${getAttributeValue(tag, "value") ?? "default"} (Click to edit)`,
+              { type: "attribute", insertionId: "mstts:audioduration", attributeName: "value", tagRange }
+            ),
+            createLens(model, tagStart, tagEnd + 1, "Delete", {
+              type: "delete",
+              tagRange,
+              elementRange
+            })
+          );
+        } else if (tagName === "break" && /\/\s*>$/.test(tag)) {
           lenses.push(
             createLens(
               model,
@@ -3813,6 +3956,17 @@ var SSML_TAG_DEFINITIONS = [
         values: VISEME_TYPE_PRESETS2
       }
     ]
+  },
+  {
+    name: "mstts:audioduration",
+    description: "Sets the target duration of synthesized audio.",
+    parameters: [
+      {
+        name: "value",
+        description: "The target duration, such as `10s`, `5000ms`, or `00:00:10`.",
+        example: "10s"
+      }
+    ]
   }
 ];
 var definitionsByName = /* @__PURE__ */ new Map();
@@ -4420,7 +4574,7 @@ function useSsmlMonaco({
 // src/SsmlEditor.tsx
 var import_jsx_runtime6 = require("react/jsx-runtime");
 var UNGROUPED_TOOLBAR_GROUP = "__ssml-editor-ungrouped__";
-var TIMING_POPOVER_TAGS = /* @__PURE__ */ new Set(["break", "mstts:silence"]);
+var TIMING_POPOVER_TAGS = /* @__PURE__ */ new Set(["break", "mstts:silence", "mstts:audioduration"]);
 var PROSODY_POPOVER_TAGS = /* @__PURE__ */ new Set(["prosody", "mstts:express-as", "voice", "emphasis"]);
 var TEXT_POPOVER_TAGS = /* @__PURE__ */ new Set(["sub", "say-as", "phoneme", "w", "lang"]);
 function localizedText(value) {
@@ -4648,13 +4802,34 @@ var SSML_INSERTIONS = [
       suffix: "",
       mode: "insert"
     })
+  },
+  {
+    id: "mstts:audioduration",
+    icon: "\u25F7",
+    tagName: "mstts:audioduration",
+    selfClosing: true,
+    labels: { ja: "\u97F3\u58F0\u9577", en: "Audio duration" },
+    descriptions: {
+      ja: "\u5408\u6210\u97F3\u58F0\u306E\u76EE\u6A19\u6642\u9593\u3092\u6307\u5B9A\u3057\u307E\u3059\u3002",
+      en: "Sets the target duration of synthesized audio."
+    },
+    parameterDescription: {
+      ja: "\u76EE\u6A19\u6642\u9593\u3092\u9078\u629E\u3057\u307E\u3059\u3002",
+      en: "Selects the target duration."
+    },
+    options: createInsertionOptions(AUDIO_DURATION_PRESETS, AUDIO_DURATION_DESCRIPTIONS),
+    createTemplate: (value) => ({
+      prefix: `<mstts:audioduration value="${value}"/>`,
+      suffix: "",
+      mode: "insert"
+    })
   }
 ];
 var DEFAULT_INSERTION_GROUPS = [
   {
     id: "pauses",
     labels: { ja: "\u9593\u30FB\u7121\u97F3", en: "Pauses" },
-    insertionIds: ["break", "mstts:silence"]
+    insertionIds: ["break", "mstts:silence", "mstts:audioduration"]
   },
   {
     id: "prosody",