From a42dd4c71258729f88942cdec473c0c9ffdc140d Mon Sep 17 00:00:00 2001
From: Xin Wang <wx44wx@XindeMac-mini.local>
Date: Mon, 9 Feb 2026 14:39:43 +0800
Subject: [PATCH] Fix tts voice bug

---
 web/pages/Assistants.tsx   | 25 ++++++++++++++++++++++++-
 web/pages/VoiceLibrary.tsx | 24 ++++++++++++++++++++----
 2 files changed, 44 insertions(+), 5 deletions(-)

diff --git a/web/pages/Assistants.tsx b/web/pages/Assistants.tsx
index 93ce445..9759930 100644
--- a/web/pages/Assistants.tsx
+++ b/web/pages/Assistants.tsx
@@ -19,6 +19,29 @@ const isSiliconflowVendor = (vendor?: string) => {
   return normalized === 'siliconflow' || normalized === '硅基流动';
 };
 
+const SILICONFLOW_DEFAULT_MODEL = 'FunAudioLLM/CosyVoice2-0.5B';
+
+const buildSiliconflowVoiceKey = (voiceId: string, model?: string) => {
+  const id = String(voiceId || '').trim();
+  if (!id) return '';
+  if (id.includes(':')) return id;
+  return `${model || SILICONFLOW_DEFAULT_MODEL}:${id}`;
+};
+
+const resolveRuntimeTtsVoice = (selectedVoiceId: string, voice: Voice) => {
+  const explicitKey = String(voice.voiceKey || '').trim();
+  if (!isSiliconflowVendor(voice.vendor)) {
+    return explicitKey || selectedVoiceId;
+  }
+  if (voice.isSystem) {
+    const canonical = buildSiliconflowVoiceKey(selectedVoiceId, voice.model);
+    if (!explicitKey) return canonical;
+    const explicitSuffix = explicitKey.includes(':') ? explicitKey.split(':').pop() : explicitKey;
+    if (explicitSuffix && explicitSuffix !== selectedVoiceId) return canonical;
+  }
+  return explicitKey || buildSiliconflowVoiceKey(selectedVoiceId, voice.model);
+};
+
 export const AssistantsPage: React.FC = () => {
   const [assistants, setAssistants] = useState<Assistant[]>([]);
   const [voices, setVoices] = useState<Voice[]>([]);
@@ -1404,7 +1427,7 @@ export const DebugDrawer: React.FC<{
           provider: ttsProvider,
           model: voice.model,
           apiKey: ttsProvider === 'siliconflow' ? voice.apiKey : null,
-          voice: voice.voiceKey || voice.id,
+          voice: resolveRuntimeTtsVoice(assistant.voice, voice),
           speed: assistant.speed || voice.speed || 1.0,
         };
       } else {
diff --git a/web/pages/VoiceLibrary.tsx b/web/pages/VoiceLibrary.tsx
index 93a13a1..ea8f234 100644
--- a/web/pages/VoiceLibrary.tsx
+++ b/web/pages/VoiceLibrary.tsx
@@ -4,6 +4,14 @@ import { Button, Input, TableHeader, TableRow, TableHead, TableCell, Dialog, Bad
 import { Voice } from '../types';
 import { createVoice, deleteVoice, fetchVoices, previewVoice, updateVoice } from '../services/backendApi';
 
+const SILICONFLOW_DEFAULT_MODEL = 'FunAudioLLM/CosyVoice2-0.5B';
+
+const buildSiliconflowVoiceKey = (rawId: string, model: string): string => {
+  const id = (rawId || '').trim();
+  if (!id) return `${model}:anna`;
+  return id.includes(':') ? id : `${model}:${id}`;
+};
+
 export const VoiceLibraryPage: React.FC = () => {
   const [voices, setVoices] = useState<Voice[]>([]);
   const [searchTerm, setSearchTerm] = useState('');
@@ -249,7 +257,7 @@ const AddVoiceModal: React.FC<{
   const [vendor, setVendor] = useState<'硅基流动' | 'Ali' | 'Volcano' | 'Minimax'>('硅基流动');
   const [name, setName] = useState('');
 
-  const [sfModel, setSfModel] = useState('FunAudioLLM/CosyVoice2-0.5B');
+  const [sfModel, setSfModel] = useState(SILICONFLOW_DEFAULT_MODEL);
   const [sfVoiceId, setSfVoiceId] = useState('FunAudioLLM/CosyVoice2-0.5B:anna');
   const [sfSpeed, setSfSpeed] = useState(1);
   const [sfGain, setSfGain] = useState(0);
@@ -271,6 +279,8 @@ const AddVoiceModal: React.FC<{
   useEffect(() => {
     if (!initialVoice) return;
     const nextVendor = initialVoice.vendor === 'SiliconFlow' ? '硅基流动' : initialVoice.vendor;
+    const nextModel = initialVoice.model || SILICONFLOW_DEFAULT_MODEL;
+    const defaultVoiceKey = buildSiliconflowVoiceKey(initialVoice.id || initialVoice.name || '', nextModel);
     setVendor((nextVendor as any) || '硅基流动');
     setName(initialVoice.name || '');
     setGender(initialVoice.gender || 'Female');
@@ -278,8 +288,8 @@ const AddVoiceModal: React.FC<{
     setDescription(initialVoice.description || '');
     setModel(initialVoice.model || '');
     setVoiceKey(initialVoice.voiceKey || '');
-    setSfModel(initialVoice.model || 'FunAudioLLM/CosyVoice2-0.5B');
-    setSfVoiceId(initialVoice.voiceKey || 'FunAudioLLM/CosyVoice2-0.5B:anna');
+    setSfModel(nextModel);
+    setSfVoiceId((initialVoice.voiceKey || '').trim() || defaultVoiceKey);
     setSfSpeed(initialVoice.speed ?? 1);
     setSfGain(initialVoice.gain ?? 0);
     setSfPitch(initialVoice.pitch ?? 0);
@@ -315,6 +325,12 @@ const AddVoiceModal: React.FC<{
       return;
     }
 
+    const resolvedSiliconflowVoiceKey = (() => {
+      const current = (sfVoiceId || '').trim();
+      if (current) return current;
+      return buildSiliconflowVoiceKey(initialVoice?.id || name, sfModel || SILICONFLOW_DEFAULT_MODEL);
+    })();
+
     const newVoice: Voice = {
       id: initialVoice?.id || `${vendor === '硅基流动' ? 'sf' : 'gen'}-${Date.now()}`,
       name,
@@ -323,7 +339,7 @@ const AddVoiceModal: React.FC<{
       language,
       description: description || (vendor === '硅基流动' ? `Model: ${sfModel}` : `Model: ${model}`),
       model: vendor === '硅基流动' ? sfModel : model,
-      voiceKey: vendor === '硅基流动' ? sfVoiceId : voiceKey,
+      voiceKey: vendor === '硅基流动' ? resolvedSiliconflowVoiceKey : voiceKey,
       apiKey,
       baseUrl,
       speed: sfSpeed,