tts: handle invalid language code and show no voices hints, closes #1579 (#1607)

apps/readest-app/public/locales/ar/translation.json

@@ -466,5 +466,6 @@
   "Open Library": "المكتبة المفتوحة",
   "Fiction, Science, History": "خيال، علوم، تاريخ",
   "Select Cover Image": "اختر صورة الغلاف",
-  "Open Book in New Window": "فتح الكتاب في نافذة جديدة"
+  "Open Book in New Window": "فتح الكتاب في نافذة جديدة",
+  "Voices for {{lang}}": "الأصوات لـ {{lang}}"
 }

apps/readest-app/public/locales/de/translation.json

@@ -450,5 +450,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Belletristik, Wissenschaft, Geschichte",
   "Select Cover Image": "Coverbild auswählen",
-  "Open Book in New Window": "Buch in neuem Fenster öffnen"
+  "Open Book in New Window": "Buch in neuem Fenster öffnen",
+  "Voices for {{lang}}": "Stimmen für {{lang}}"
 }

apps/readest-app/public/locales/el/translation.json

@@ -450,5 +450,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Μυθοπλασία, Επιστήμη, Ιστορία",
   "Select Cover Image": "Επιλέξτε εικόνα εξωφύλλου",
-  "Open Book in New Window": "Άνοιγμα βιβλίου σε νέο παράθυρο"
+  "Open Book in New Window": "Άνοιγμα βιβλίου σε νέο παράθυρο",
+  "Voices for {{lang}}": "Φωνές για {{lang}}"
 }

apps/readest-app/public/locales/es/translation.json

@@ -454,5 +454,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Ficción, Ciencia, Historia",
   "Select Cover Image": "Seleccionar imagen de portada",
-  "Open Book in New Window": "Abre el libro en una nueva ventana"
+  "Open Book in New Window": "Abre el libro en una nueva ventana",
+  "Voices for {{lang}}": "Voces para {{lang}}"
 }

apps/readest-app/public/locales/fr/translation.json

@@ -454,5 +454,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Fiction, Science, Histoire",
   "Select Cover Image": "Sélectionner l'image de couverture",
-  "Open Book in New Window": "Ouvrir dans une nouvelle fenêtre"
+  "Open Book in New Window": "Ouvrir dans une nouvelle fenêtre",
+  "Voices for {{lang}}": "Voix pour {{lang}}"
 }

apps/readest-app/public/locales/hi/translation.json

@@ -450,5 +450,6 @@
   "Open Library": "ओपन लाइब्रेरी",
   "Fiction, Science, History": "काल्पनिक, विज्ञान, इतिहास",
   "Select Cover Image": "कवर इमेज चुनें",
-  "Open Book in New Window": "नई विंडो में पुस्तक खोलें"
+  "Open Book in New Window": "नई विंडो में पुस्तक खोलें",
+  "Voices for {{lang}}": "{{lang}} के लिए आवाज़ें"
 }

apps/readest-app/public/locales/id/translation.json

@@ -446,5 +446,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Fiksi, Sains, Sejarah",
   "Select Cover Image": "Pilihan Gambar Sampul",
-  "Open Book in New Window": "Buka Buku di Jendela Baru"
+  "Open Book in New Window": "Buka Buku di Jendela Baru",
+  "Voices for {{lang}}": "Suara untuk {{lang}}"
 }

apps/readest-app/public/locales/it/translation.json

@@ -454,5 +454,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Narrativa, Scienza, Storia",
   "Select Cover Image": "Seleziona immagine di copertina",
-  "Open Book in New Window": "Apri libro in una nuova finestra"
+  "Open Book in New Window": "Apri libro in una nuova finestra",
+  "Voices for {{lang}}": "Voci per {{lang}}"
 }

apps/readest-app/public/locales/ja/translation.json

@@ -446,5 +446,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "小説、科学、歴史",
   "Select Cover Image": "表紙画像を選択",
-  "Open Book in New Window": "新しいウィンドウで本を開く"
+  "Open Book in New Window": "新しいウィンドウで本を開く",
+  "Voices for {{lang}}": "{{lang}}の音声"
 }

apps/readest-app/public/locales/ko/translation.json

@@ -446,5 +446,6 @@
   "Open Library": "오픈 라이브러리",
   "Fiction, Science, History": "소설, 과학, 역사",
   "Select Cover Image": "표지 이미지 선택",
-  "Open Book in New Window": "새 창에서 책 열기"
+  "Open Book in New Window": "새 창에서 책 열기",
+  "Voices for {{lang}}": "{{lang}}의 음성"
 }

apps/readest-app/public/locales/nl/translation.json

@@ -450,5 +450,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Fictie, Wetenschap, Geschiedenis",
   "Select Cover Image": "Selecteer omslagafbeelding",
-  "Open Book in New Window": "Open boek in nieuw venster"
+  "Open Book in New Window": "Open boek in nieuw venster",
+  "Voices for {{lang}}": "Stemmen voor {{lang}}"
 }

apps/readest-app/public/locales/pl/translation.json

@@ -458,5 +458,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Beletrystyka, Nauka, Historia",
   "Select Cover Image": "Zaznacz obraz okładki",
-  "Open Book in New Window": "Otwórz książkę w nowym oknie"
+  "Open Book in New Window": "Otwórz książkę w nowym oknie",
+  "Voices for {{lang}}": "{{lang}} Głosy"
 }

apps/readest-app/public/locales/pt/translation.json

@@ -454,5 +454,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Ficção, Ciência, História",
   "Select Cover Image": "Selecionar imagem da capa",
-  "Open Book in New Window": "Abrir livro em nova janela"
+  "Open Book in New Window": "Abrir livro em nova janela",
+  "Voices for {{lang}}": "Vozes para {{lang}}"
 }

apps/readest-app/public/locales/ru/translation.json

@@ -458,5 +458,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Художественная литература, Наука, История",
   "Select Cover Image": "Выбрать изображение обложки",
-  "Open Book in New Window": "Открыть книгу в новом окне"
+  "Open Book in New Window": "Открыть книгу в новом окне",
+  "Voices for {{lang}}": "Голоса для {{lang}}"
 }

apps/readest-app/public/locales/th/translation.json

@@ -446,5 +446,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "นิยาย, วิทยาศาสตร์, ประวัติศาสตร์",
   "Select Cover Image": "เลือกภาพปก",
-  "Open Book in New Window": "เปิดหนังสือในหน้าต่างใหม่"
+  "Open Book in New Window": "เปิดหนังสือในหน้าต่างใหม่",
+  "Voices for {{lang}}": "เสียงสำหรับ {{lang}}"
 }

apps/readest-app/public/locales/tr/translation.json

@@ -450,5 +450,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Kurgu, Bilim, Tarih",
   "Select Cover Image": "Kapak Resmi Seç",
-  "Open Book in New Window": "Kitabı Yeni Pencerede Aç"
+  "Open Book in New Window": "Kitabı Yeni Pencerede Aç",
+  "Voices for {{lang}}": "{{lang}} için Sesler"
 }

apps/readest-app/public/locales/uk/translation.json

@@ -458,5 +458,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Художня література, Наукова фантастика, Історія",
   "Select Cover Image": "Оберіть зображення обкладинки",
-  "Open Book in New Window": "Відкрити книгу в новому вікні"
+  "Open Book in New Window": "Відкрити книгу в новому вікні",
+  "Voices for {{lang}}": "Голоси для {{lang}}"
 }

apps/readest-app/public/locales/vi/translation.json

@@ -446,5 +446,6 @@
   "Open Library": "Open Library",
   "Fiction, Science, History": "Tiểu thuyết, Khoa học, Lịch sử",
   "Select Cover Image": "Chọn ảnh bìa",
-  "Open Book in New Window": "Open Sách trong Cửa sổ Mới"
+  "Open Book in New Window": "Open Sách trong Cửa sổ Mới",
+  "Voices for {{lang}}": "Giọng nói cho {{lang}}"
 }

apps/readest-app/public/locales/zh-CN/translation.json

@@ -317,7 +317,7 @@
   "Quota Exceeded": "额度不足",
   "{{percentage}}% of Daily Translation Characters Used.": "已使用 {{percentage}}% 的每日翻译字符数",
   "Translation Characters": "翻译字数",
-  "{{engine}}: {{count}} voices_other": "{{engine}}: {{count}} 种声音",
+  "{{engine}}: {{count}} voices_other": "{{engine}}: {{count}} 种语音",
   "Something went wrong. Don't worry, our team has been notified and we're working on a fix.": "出了点问题。请不要担心，我们的团队已经收到通知并正在修复中。",
   "Error Details:": "错误详情：",
   "Try Again": "重试",
@@ -446,5 +446,6 @@
   "Open Library": "开放图书馆",
   "Fiction, Science, History": "小说，科学，历史",
   "Select Cover Image": "选择封面图片",
-  "Open Book in New Window": "在新窗口中打开书籍"
+  "Open Book in New Window": "在新窗口中打开书籍",
+  "Voices for {{lang}}": "{{lang}} 的语音"
 }

apps/readest-app/public/locales/zh-TW/translation.json

@@ -446,5 +446,6 @@
   "Open Library": "開放圖書館",
   "Fiction, Science, History": "小說，科學，歷史",
   "Select Cover Image": "選擇封面圖片",
-  "Open Book in New Window": "在新視窗中打開書籍"
+  "Open Book in New Window": "在新視窗中打開書籍",
+  "Voices for {{lang}}": "{{lang}} 的語音"
 }

apps/readest-app/src/app/reader/components/tts/TTSControl.tsx

@@ -212,11 +212,7 @@ const TTSControl: React.FC<TTSControlProps> = ({ bookKey }) => {
       await ttsController.initViewTTS();
       const ssml = view.tts?.from(ttsFromRange);
       if (ssml) {
-        let lang = parseSSMLLang(ssml) || 'en';
-        // We will not trust 'en' language from ssml, as it may be a fallback or hardcoded value
-        if (lang === 'en' && primaryLang && primaryLang !== 'en') {
-          lang = primaryLang.split('-')[0]!;
-        }
+        let lang = parseSSMLLang(ssml, primaryLang) || 'en';
         setIsPlaying(true);
         setTtsLang(lang);
 

apps/readest-app/src/app/reader/components/tts/TTSPanel.tsx

@@ -9,6 +9,7 @@ import { useReaderStore } from '@/store/readerStore';
 import { TranslationFunc, useTranslation } from '@/hooks/useTranslation';
 import { useSettingsStore } from '@/store/settingsStore';
 import { useDefaultIconSize, useResponsiveSize } from '@/hooks/useResponsiveSize';
+import { getLanguageName } from '@/utils/lang';
 
 type TTSPanelProps = {
   bookKey: string;
@@ -178,7 +179,19 @@ const TTSPanel = ({
   useEffect(() => {
     const fetchVoices = async () => {
       const voiceGroups = await onGetVoices(ttsLang);
-      setVoiceGroups(voiceGroups);
+      const voicesCount = voiceGroups.reduce((acc, group) => acc + group.voices.length, 0);
+      if (!voiceGroups || voicesCount === 0) {
+        console.warn('No voices found for TTSPanel');
+        setVoiceGroups([
+          {
+            id: 'no-voices',
+            name: _('Voices for {{lang}}', { lang: getLanguageName(ttsLang) }),
+            voices: [],
+          },
+        ]);
+      } else {
+        setVoiceGroups(voiceGroups);
+      }
     };
     fetchVoices();
     // eslint-disable-next-line react-hooks/exhaustive-deps
@@ -289,48 +302,44 @@ const TTSPanel = ({
           >
             {voiceGroups.map((voiceGroup, index) => {
               return (
-                voiceGroup.voices.length > 0 && (
-                  <div key={voiceGroup.id} className=''>
-                    <div className='flex items-center gap-2 px-2 py-1'>
-                      <span
-                        style={{ width: `${defaultIconSize}px`, height: `${defaultIconSize}px` }}
-                      ></span>
-                      <span className='text-sm text-gray-400 sm:text-xs'>
-                        {_('{{engine}}: {{count}} voices', {
-                          engine: _(voiceGroup.name),
-                          count: voiceGroup.voices.length,
-                        })}
-                      </span>
-                    </div>
-                    {voiceGroup.voices.map((voice, voiceIndex) => (
-                      <li
-                        key={`${index}-${voiceGroup.id}-${voiceIndex}`}
-                        onClick={() => !voice.disabled && handleSelectVoice(voice.id, voice.lang)}
-                      >
-                        <div className='flex items-center px-2'>
-                          <span
-                            style={{
-                              width: `${defaultIconSize}px`,
-                              height: `${defaultIconSize}px`,
-                            }}
-                          >
-                            {selectedVoice === voice.id && (
-                              <MdCheck className='text-base-content' />
-                            )}
-                          </span>
-                          <span
-                            className={clsx(
-                              'text-base sm:text-sm',
-                              voice.disabled && 'text-gray-400',
-                            )}
-                          >
-                            {_(voice.name)}
-                          </span>
-                        </div>
-                      </li>
-                    ))}
+                <div key={voiceGroup.id} className=''>
+                  <div className='flex items-center gap-2 px-2 py-1'>
+                    <span
+                      style={{ width: `${defaultIconSize}px`, height: `${defaultIconSize}px` }}
+                    ></span>
+                    <span className='text-sm text-gray-400 sm:text-xs'>
+                      {_('{{engine}}: {{count}} voices', {
+                        engine: _(voiceGroup.name),
+                        count: voiceGroup.voices.length,
+                      })}
+                    </span>
                   </div>
-                )
+                  {voiceGroup.voices.map((voice, voiceIndex) => (
+                    <li
+                      key={`${index}-${voiceGroup.id}-${voiceIndex}`}
+                      onClick={() => !voice.disabled && handleSelectVoice(voice.id, voice.lang)}
+                    >
+                      <div className='flex items-center px-2'>
+                        <span
+                          style={{
+                            width: `${defaultIconSize}px`,
+                            height: `${defaultIconSize}px`,
+                          }}
+                        >
+                          {selectedVoice === voice.id && <MdCheck className='text-base-content' />}
+                        </span>
+                        <span
+                          className={clsx(
+                            'text-base sm:text-sm',
+                            voice.disabled && 'text-gray-400',
+                          )}
+                        >
+                          {_(voice.name)}
+                        </span>
+                      </div>
+                    </li>
+                  ))}
+                </div>
               );
             })}
           </ul>

apps/readest-app/src/services/tts/EdgeTTSClient.ts

@@ -1,7 +1,7 @@
 import { getUserLocale } from '@/utils/misc';
 import { TTSClient, TTSMessageEvent } from './TTSClient';
 import { EdgeSpeechTTS, EdgeTTSPayload } from '@/libs/edgeTTS';
-import { parseSSMLLang, parseSSMLMarks } from '@/utils/ssml';
+import { parseSSMLMarks } from '@/utils/ssml';
 import { TTSController } from './TTSController';
 import { TTSUtils } from './TTSUtils';
 import { TTSGranularity, TTSVoice, TTSVoicesGroup } from './types';
@@ -60,19 +60,14 @@ export class EdgeTTSClient implements TTSClient {
   };
 
   async *speak(ssml: string, signal: AbortSignal, preload = false) {
-    const { marks } = parseSSMLMarks(ssml);
-    let defaultLang = parseSSMLLang(ssml) || 'en';
-    if (defaultLang === 'en' && this.#primaryLang && this.#primaryLang !== 'en') {
-      defaultLang = this.#primaryLang;
-    }
+    const { marks } = parseSSMLMarks(ssml, this.#primaryLang);
 
     if (preload) {
       // preload the first 2 marks immediately and the rest in the background
       const maxImmediate = 2;
       for (let i = 0; i < Math.min(maxImmediate, marks.length); i++) {
         const mark = marks[i]!;
-        const { language } = mark;
-        const voiceLang = language || defaultLang;
+        const { language: voiceLang } = mark;
         const voiceId = await this.getVoiceIdFromLang(voiceLang);
         this.#currentVoiceId = voiceId;
         await this.#edgeTTS
@@ -86,8 +81,7 @@ export class EdgeTTSClient implements TTSClient {
           for (let i = maxImmediate; i < marks.length; i++) {
             const mark = marks[i]!;
             try {
-              const { language } = mark;
-              const voiceLang = language || defaultLang;
+              const { language: voiceLang } = mark;
               const voiceId = await this.getVoiceIdFromLang(voiceLang);
               await this.#edgeTTS.createAudio(this.getPayload(voiceLang, mark.text, voiceId));
             } catch (err) {
@@ -116,8 +110,7 @@ export class EdgeTTSClient implements TTSClient {
         break;
       }
       try {
-        const { language } = mark;
-        const voiceLang = language || defaultLang;
+        const { language: voiceLang } = mark;
         const voiceId = await this.getVoiceIdFromLang(voiceLang);
         this.#speakingLang = voiceLang;
         const blob = await this.#edgeTTS.createAudio(

apps/readest-app/src/services/tts/NativeTTSClient.ts

@@ -1,7 +1,7 @@
 import { invoke } from '@tauri-apps/api/core';
 import { addPluginListener, PluginListener } from '@tauri-apps/api/core';
 import { getUserLocale } from '@/utils/misc';
-import { parseSSMLLang, parseSSMLMarks } from '@/utils/ssml';
+import { parseSSMLMarks } from '@/utils/ssml';
 import { stubTranslation as _ } from '@/utils/misc';
 import { TTSClient, TTSMessageEvent } from './TTSClient';
 import { TTSGranularity, TTSMark, TTSVoice, TTSVoicesGroup } from './types';
@@ -100,13 +100,12 @@ export class NativeTTSClient implements TTSClient {
     return this.initialized;
   }
 
-  async *speakMark(mark: TTSMark, preload: boolean, defaultLang: string, signal: AbortSignal) {
+  async *speakMark(mark: TTSMark, preload: boolean, signal: AbortSignal) {
     if (preload) {
       yield { code: 'end', message: 'Dummy preload finished' } as TTSMessageEvent;
       return;
     }
-    const { language } = mark;
-    const voiceLang = language || defaultLang;
+    const { language: voiceLang } = mark;
     const voiceId = await this.getVoiceIdFromLang(voiceLang);
     this.#currentVoiceId = voiceId;
     this.#speakingLang = voiceLang;
@@ -179,14 +178,10 @@ export class NativeTTSClient implements TTSClient {
   }
 
   async *speak(ssml: string, signal: AbortSignal, preload: boolean = false) {
-    const { marks } = parseSSMLMarks(ssml);
-    let defaultLang = parseSSMLLang(ssml) || 'en';
-    if (defaultLang === 'en' && this.#primaryLang && this.#primaryLang !== 'en') {
-      defaultLang = this.#primaryLang;
-    }
+    const { marks } = parseSSMLMarks(ssml, this.#primaryLang);
 
     for (const mark of marks) {
-      for await (const ev of this.speakMark(mark, preload, defaultLang, signal)) {
+      for await (const ev of this.speakMark(mark, preload, signal)) {
         if (signal.aborted) {
           yield { code: 'error', message: 'Aborted' } as TTSMessageEvent;
           return;

apps/readest-app/src/services/tts/WebSpeechClient.ts

@@ -1,6 +1,6 @@
 import { getUserLocale } from '@/utils/misc';
 import { TTSClient, TTSMessageEvent } from './TTSClient';
-import { parseSSMLLang, parseSSMLMarks } from '@/utils/ssml';
+import { parseSSMLMarks } from '@/utils/ssml';
 import { TTSGranularity, TTSMark, TTSVoice, TTSVoicesGroup } from './types';
 import { WEB_SPEECH_BLACKLISTED_VOICES } from './TTSData';
 import { TTSController } from './TTSController';
@@ -18,7 +18,7 @@ interface TTSBoundaryEvent {
 
 async function* speakWithMarks(
   ssml: string,
-  defaultLang: string,
+  primaryLang: string,
   getRate: () => number,
   getPitch: () => number,
   getVoice: (lang: string) => Promise<SpeechSynthesisVoice | null>,
@@ -26,14 +26,11 @@ async function* speakWithMarks(
   setSpeakingLang: (lang: string) => void,
   dispatchSpeakMark: (mark: TTSMark) => void,
 ) {
-  const { marks } = parseSSMLMarks(ssml);
-
+  const { marks } = parseSSMLMarks(ssml, primaryLang);
   const synth = window.speechSynthesis;
-
   const utterance = new SpeechSynthesisUtterance();
   for (const mark of marks) {
-    const { language } = mark;
-    const voiceLang = language || defaultLang;
+    const { language: voiceLang } = mark;
     dispatchSpeakMark(mark);
     utterance.text = mark.text;
     utterance.rate = getRate();
@@ -149,14 +146,9 @@ export class WebSpeechClient implements TTSClient {
     // no need to preload for web speech
     if (preload) return;
 
-    let defaultLang = parseSSMLLang(ssml) || 'en';
-    if (defaultLang === 'en' && this.#primaryLang && this.#primaryLang !== 'en') {
-      defaultLang = this.#primaryLang;
-    }
-
     for await (const ev of speakWithMarks(
       ssml,
-      defaultLang,
+      this.#primaryLang,
       () => this.#rate,
       () => this.#pitch,
       this.getWebSpeechVoiceFromLang,

apps/readest-app/src/services/tts/types.ts

@@ -23,5 +23,5 @@ export type TTSMark = {
   offset: number;
   name: string;
   text: string;
-  language?: string;
+  language: string;
 };

apps/readest-app/src/utils/lang.ts

@@ -68,7 +68,32 @@ export const isSameLang = (lang1?: string | null, lang2?: string | null): boolea
   return normalizedLang1 === normalizedLang2;
 };
 
+export const isValidLang = (lang?: string) => {
+  if (!lang) return false;
+  const code = lang.split('-')[0]!.toLowerCase();
+  return iso6392.some((l) => l.iso6391 === code || l.iso6392B === code);
+};
+
 export const code6392to6391 = (code: string): string => {
   const lang = iso6392.find((l) => l.iso6392B === code);
   return lang?.iso6391 || '';
 };
+
+export const getLanguageName = (code: string): string => {
+  const lang = normalizedLangCode(code);
+  const language = iso6392.find((l) => l.iso6391 === lang || l.iso6392B === lang);
+  return language ? language.name : lang;
+};
+
+export const inferLangFromScript = (text: string, lang: string): string => {
+  if (!lang || lang === 'en') {
+    if (/[\p{Script=Hangul}]/u.test(text)) {
+      return 'ko';
+    } else if (/[\p{Script=Hiragana}\p{Script=Katakana}]/u.test(text)) {
+      return 'ja';
+    } else if (/[\p{Script=Han}]/u.test(text)) {
+      return 'zh';
+    }
+  }
+  return lang;
+};

apps/readest-app/src/utils/ssml.ts

@@ -1,4 +1,5 @@
 import { TTSMark } from '@/services/tts/types';
+import { inferLangFromScript, isSameLang, isValidLang } from './lang';
 
 const cleanTextContent = (text: string) =>
   text.replace(/\r\n/g, '  ').replace(/\r/g, ' ').replace(/\n/g, ' ').trimStart();
@@ -16,9 +17,28 @@ export const genSSML = (lang: string, text: string, voice: string, rate: number)
   `;
 };
 
-export const parseSSMLMarks = (ssml: string) => {
-  const speakMatch = ssml.match(/<speak[^>]*xml:lang="([^"]+)"[^>]*>/i);
-  const defaultLang = speakMatch?.[1] ?? 'en';
+export const parseSSMLLang = (ssml: string, primaryLang?: string): string => {
+  let lang = 'en';
+  const match = ssml.match(/xml:lang\s*=\s*"([^"]+)"/);
+  if (match && match[1]) {
+    const parts = match[1].split('-');
+    lang =
+      parts.length > 1
+        ? `${parts[0]!.toLowerCase()}-${parts[1]!.toUpperCase()}`
+        : parts[0]!.toLowerCase();
+
+    if (!isValidLang(lang)) {
+      lang = 'en';
+    }
+  }
+  if (lang === 'en' && primaryLang && !isSameLang(lang, primaryLang)) {
+    lang = primaryLang.split('-')[0]!.toLowerCase();
+  }
+  return inferLangFromScript(ssml, lang);
+};
+
+export const parseSSMLMarks = (ssml: string, primaryLang?: string) => {
+  const defaultLang = parseSSMLLang(ssml, primaryLang) || 'en';
   ssml = ssml.replace(/<speak[^>]*>/i, '').replace(/<\/speak>/i, '');
 
   let plainText = '';
@@ -93,29 +113,3 @@ export const findSSMLMark = (charIndex: number, marks: TTSMark[]) => {
 
   return result;
 };
-
-const inferLangFromScript = (text: string, lang: string | null): string | null => {
-  if (!lang || lang.startsWith('en')) {
-    if (/[\p{Script=Hangul}]/u.test(text)) {
-      return 'ko';
-    } else if (/[\p{Script=Hiragana}\p{Script=Katakana}]/u.test(text)) {
-      return 'ja';
-    } else if (/[\p{Script=Han}]/u.test(text)) {
-      return 'zh';
-    }
-  }
-  return lang;
-};
-
-export const parseSSMLLang = (ssml: string): string | null => {
-  let lang = null;
-  const match = ssml.match(/xml:lang\s*=\s*"([^"]+)"/);
-  if (match && match[1]) {
-    const parts = match[1].split('-');
-    lang =
-      parts.length > 1
-        ? `${parts[0]!.toLowerCase()}-${parts[1]!.toUpperCase()}`
-        : parts[0]!.toLowerCase();
-  }
-  return inferLangFromScript(ssml, lang);
-};