From 1c6cc2bca00e092aa20a5be5e883d51b1250e430 Mon Sep 17 00:00:00 2001
From: Huang Xin <chrox.huang@gmail.com>
Date: Tue, 4 Feb 2025 23:44:59 +0100
Subject: [PATCH] tts: change voice language at speak time when the book
 language is not accurate (#296)

---
 apps/readest-app/src/services/tts/EdgeTTSClient.ts  | 13 +++++++++++--
 .../readest-app/src/services/tts/WebSpeechClient.ts | 13 +++++++++++--
 2 files changed, 22 insertions(+), 4 deletions(-)

diff --git a/apps/readest-app/src/services/tts/EdgeTTSClient.ts b/apps/readest-app/src/services/tts/EdgeTTSClient.ts
index ca5ed0fe..d2ec1d10 100644
--- a/apps/readest-app/src/services/tts/EdgeTTSClient.ts
+++ b/apps/readest-app/src/services/tts/EdgeTTSClient.ts
@@ -8,6 +8,7 @@ export class EdgeTTSClient implements TTSClient {
   #rate = 1.0;
   #pitch = 1.0;
   #voice: TTSVoice | null = null;
+  #currentVoiceLang = '';
   #voices: TTSVoice[] = [];
   #edgeTTS: EdgeSpeechTTS;
 
@@ -60,12 +61,17 @@ export class EdgeTTSClient implements TTSClient {
     preload = false,
   ): AsyncGenerator<TTSMessageEvent> {
     const { marks } = parseSSMLMarks(ssml);
-    const lang = parseSSMLLang(ssml) || 'en';
+    const ssmlLang = parseSSMLLang(ssml) || 'en';
+    let lang = ssmlLang;
+    if (lang === 'en' && /[\p{Script=Han}]/u.test(ssml)) {
+      lang = 'zh';
+    }
 
     let voiceId = 'en-US-AriaNeural';
-    if (!this.#voice) {
+    if (!this.#voice || ssmlLang !== lang) {
       const voices = await this.getVoices(lang);
       this.#voice = voices[0] ? voices[0] : this.#voices.find((v) => v.id === voiceId) || null;
+      this.#currentVoiceLang = lang;
     }
     if (this.#voice) {
       voiceId = this.#voice.id;
@@ -229,6 +235,9 @@ export class EdgeTTSClient implements TTSClient {
   }
 
   async getVoices(lang: string): Promise<TTSVoice[]> {
+    if (this.#currentVoiceLang) {
+      lang = this.#currentVoiceLang;
+    }
     const locale = lang === 'en' ? getUserLocale(lang) || lang : lang;
     const voices = await this.getAllVoices();
     return voices.filter((v) => v.lang.startsWith(locale));
diff --git a/apps/readest-app/src/services/tts/WebSpeechClient.ts b/apps/readest-app/src/services/tts/WebSpeechClient.ts
index 93bb98f9..c144eaba 100644
--- a/apps/readest-app/src/services/tts/WebSpeechClient.ts
+++ b/apps/readest-app/src/services/tts/WebSpeechClient.ts
@@ -175,6 +175,7 @@ export class WebSpeechClient implements TTSClient {
   #rate = 1.0;
   #pitch = 1.0;
   #voice: SpeechSynthesisVoice | null = null;
+  #currentVoiceLang = '';
   #voices: SpeechSynthesisVoice[] = [];
   #synth = window.speechSynthesis;
   available = true;
@@ -213,11 +214,16 @@ export class WebSpeechClient implements TTSClient {
     // no need to preload for web speech
     if (preload) return;
 
-    const lang = parseSSMLLang(ssml) || 'en';
-    if (!this.#voice) {
+    const ssmlLang = parseSSMLLang(ssml) || 'en';
+    let lang = ssmlLang;
+    if (lang === 'en' && /[\p{Script=Han}]/u.test(ssml)) {
+      lang = 'zh';
+    }
+    if (!this.#voice || ssmlLang !== lang) {
       const voices = await this.getVoices(lang);
       const voiceId = voices[0]?.id ?? '';
       this.#voice = this.#voices.find((v) => v.voiceURI === voiceId) || null;
+      this.#currentVoiceLang = lang;
     }
     for await (const ev of speakWithMarks(
       ssml,
@@ -286,6 +292,9 @@ export class WebSpeechClient implements TTSClient {
   }
 
   async getVoices(lang: string) {
+    if (this.#currentVoiceLang) {
+      lang = this.#currentVoiceLang;
+    }
     const locale = lang === 'en' ? getUserLocale(lang) || lang : lang;
     const isValidVoice = (id: string) => {
       return !id.includes('com.apple') || id.includes('com.apple.voice.compact');