fix(gemini-tts): GSV 启用时仍走 Gemini 原生 + 别名/未知 voice_id 保底 + free_voices 去重

LyaQanYi · claude · LyaQanYi · commit d0486aeec64d · 2026-05-07T20:54:58.000+08:00
回应三轮 review：

1. codex P2 (core.py:2314) — 当 core=gemini + 选了 Gemini 原生声线 + 用户启用了
   GPT-SoVITS / local CosyVoice 时，_has_custom_tts 短路 False 不够：
   get_tts_worker 会先命中 tts_custom is_custom + gsv_enabled 分支返回
   gptsovits_tts_worker / local_cosyvoice_worker，绕过 Gemini。修法是在
   get_tts_worker 的 GSV 检查之前加一个 has_custom_voice=False + core=gemini +
   is_gemini_tts_voice(voice_id) 短路，直接返回 gemini_tts_worker。

2. coderabbit Major (character_card_manager.js:5523) — Gemini 原生音色下拉只在
   voiceId === currentVoiceId 时 select，没有 GSV 那种 ensureFallback 兜底；
   一旦角色保存的是别名（如"中文男"）或本轮 catalog 没暴露该 ID，select 会
   回到首项，下次保存表单走 unregister_voice 把用户音色丢掉。仿
   _loadPanelGsvVoices 的兜底，给未知值补一条 "(?)" 占位 option 保留原值。
   配套加 character.savedVoiceFallback i18n key（8 个 locale），无 key 时回退
   到中文字面量。

3. coderabbit Minor (voice_clone.js:1264) — 原生音色去重只看自定义注册音色，
   没扣 free_voices；冲突时同 ID 会被渲两遍，多重 selected 视觉态。把
   free_voices 的 voiceId 一并加入去重 set。同样的扩展也应用到
   character_card_manager.js 的 optgroup 渲染分支。

测试：tests/unit/test_gemini_tts_voices.py 29 用例通过；node --check 两份 JS、
json.load 8 份 locale 全部通过。

Co-Authored-By: Claude Opus 4.7 (1M context) &lt;noreply@anthropic.com&gt;
diff --git a/main_logic/tts_client.py b/main_logic/tts_client.py
@@ -20,6 +20,7 @@
 from utils.config_manager import get_config_manager
 from utils.gemini_tts_voices import (
     GEMINI_TTS_MODEL,
+    is_gemini_tts_voice,
     normalize_gemini_tts_voice,
 )
 from utils.logger_config import get_module_logger
@@ -2887,6 +2888,17 @@ def get_tts_worker(core_api_type='qwen', has_custom_voice=False, voice_id=''):
             worker = partial(minimax_tts_worker, base_url=base_url)
             return worker, api_key, 'minimax'
 
+    # core=gemini + 选了 Gemini 原生声线 (Puck/Leda/中文男 等) 时优先走 Gemini，
+    # 不能被 has_custom_voice=False 的 GPT-SoVITS / local CosyVoice fallthrough 拦截 ——
+    # _has_custom_tts 已经判断 voice_id 不是用户克隆音色，这里 has_custom_voice 必为 False，
+    # 是用户显式选择的 Gemini 原生路径，应当尊重该选择喵。
+    if (
+        not has_custom_voice
+        and core_api_type == 'gemini'
+        and is_gemini_tts_voice(voice_id)
+    ):
+        return gemini_tts_worker, None, 'gemini'
+
     try:
         tts_config = cm.get_model_api_config('tts_custom')
         if tts_config.get('is_custom'):
diff --git a/static/js/character_card_manager.js b/static/js/character_card_manager.js
@@ -5501,12 +5501,18 @@ async function _loadPanelVoices(selectEl, currentVoiceId) {
             }
 
             // Gemini 原生音色（仅在 CORE_API_TYPE=gemini 时由后端注入）
-            // 与已注册自定义音色冲突时（如用户克隆了名为 Puck 的音色），优先保留自定义条目，
-            // 与 _has_custom_tts 路由优先级保持一致。
+            // 去重范围：已注册自定义音色 + 已渲染的免费预设音色 ID，
+            // 避免任一冲突时下拉里重复条目和多重 selected 视觉态。
+            // 自定义/免费音色优先保留，与 _has_custom_tts 的路由优先级一致。
             if (data.native_voices && Object.keys(data.native_voices).length > 0) {
-                const registeredVoiceIds = data.voices ? new Set(Object.keys(data.voices)) : new Set();
+                const renderedVoiceIds = new Set(data.voices ? Object.keys(data.voices) : []);
+                if (data.free_voices) {
+                    Object.values(data.free_voices).forEach(function (id) {
+                        if (id) renderedVoiceIds.add(String(id));
+                    });
+                }
                 const nativeEntries = Object.entries(data.native_voices)
-                    .filter(function ([voiceId]) { return !registeredVoiceIds.has(voiceId); });
+                    .filter(function ([voiceId]) { return !renderedVoiceIds.has(voiceId); });
                 if (nativeEntries.length > 0) {
                     const nativeGroup = document.createElement('optgroup');
                     const nativeLabel = window.t ? window.t('character.geminiNativeVoices') : 'Gemini 原生音色';
@@ -5521,6 +5527,25 @@ async function _loadPanelVoices(selectEl, currentVoiceId) {
                     });
                     selectEl.appendChild(nativeGroup);
                 }
+
+                // 保底：currentVoiceId 是 Gemini 别名（"中文男"、"male" 等）或本轮 catalog 没暴露
+                // 该 ID 时，下拉里没有匹配项 select 会回到首项；下次保存表单会被误判为
+                // "已清空"走 unregister_voice 分支，把用户保存的音色丢掉。这里仿 GSV 兜底，
+                // 给未知值补一条 "(?)" 占位条，保留原值供后端 normalize。
+                if (currentVoiceId
+                    && !selectEl.querySelector('option[value="' + CSS.escape(currentVoiceId) + '"]')) {
+                    const fallbackGroup = document.createElement('optgroup');
+                    const fallbackLabel = window.t ? window.t('character.savedVoiceFallback') : '当前已保存音色';
+                    fallbackGroup.label = '── ' + fallbackLabel + ' ──';
+                    fallbackGroup.dataset.geminiFallbackGroup = 'true';
+                    const fallbackOption = document.createElement('option');
+                    fallbackOption.value = currentVoiceId;
+                    fallbackOption.textContent = currentVoiceId + ' (?)';
+                    fallbackOption.title = currentVoiceId;
+                    fallbackOption.selected = true;
+                    fallbackGroup.appendChild(fallbackOption);
+                    selectEl.appendChild(fallbackGroup);
+                }
             }
         }
 
diff --git a/static/js/voice_clone.js b/static/js/voice_clone.js
@@ -1255,12 +1255,17 @@ async function loadVoices() {
         }
 
         // 渲染 Gemini 原生音色（CORE_API_TYPE=gemini 时由后端注入）
-        // 与已注册自定义音色冲突时（如用户克隆了名为 Puck 的音色），优先保留自定义条目，
-        // 与 _has_custom_tts 路由优先级保持一致。
+        // 去重范围：自定义注册音色 + 免费预设音色 ID，避免冲突时列表里重复条目和多重选中态。
+        // 自定义/免费音色优先保留，与 _has_custom_tts 的路由优先级一致。
         if (data.native_voices && Object.keys(data.native_voices).length > 0) {
-            const registeredVoiceIds = new Set(voicesArray.map((v) => v.voiceId));
+            const renderedVoiceIds = new Set(voicesArray.map((v) => v.voiceId));
+            if (data.free_voices) {
+                Object.values(data.free_voices).forEach((id) => {
+                    if (id) renderedVoiceIds.add(String(id));
+                });
+            }
             const nativeEntries = Object.entries(data.native_voices)
-                .filter(([voiceId]) => !registeredVoiceIds.has(voiceId));
+                .filter(([voiceId]) => !renderedVoiceIds.has(voiceId));
             if (nativeEntries.length > 0) {
                 const hasPriorContent = voicesArray.length > 0
                     || (data.free_voices && Object.keys(data.free_voices).length > 0);
diff --git a/static/locales/en.json b/static/locales/en.json
@@ -1861,6 +1861,7 @@
         "switchError": "An error occurred while switching neko",
         "freePresetVoices": "Free Preset Voices",
         "geminiNativeVoices": "Gemini Native Voices",
+        "savedVoiceFallback": "Currently saved voice",
         "gptsovitsVoices": "GPT-SoVITS Voices",
         "loadFailed": "Failed to load character data",
         "renameError": "Rename failed: {{error}}",
diff --git a/static/locales/es.json b/static/locales/es.json
@@ -1861,6 +1861,7 @@
         "switchError": "Se produjo un error al cambiar de Neko.",
         "freePresetVoices": "Voces preestablecidas gratuitas",
         "geminiNativeVoices": "Voces nativas de Gemini",
+        "savedVoiceFallback": "Voz actualmente guardada",
         "gptsovitsVoices": "Voces GPT-SoVITS",
         "loadFailed": "No se pudieron cargar los datos del personaje",
         "renameError": "Error al cambiar el nombre: {{error}}",
diff --git a/static/locales/ja.json b/static/locales/ja.json
@@ -1861,6 +1861,7 @@
         "switchError": "neko の切り替え中にエラーが発生しました",
         "freePresetVoices": "── 無料プリセット音色 ──",
         "geminiNativeVoices": "Gemini ネイティブ音声",
+        "savedVoiceFallback": "現在保存中の音声",
         "gptsovitsVoices": "GPT-SoVITS 音声",
         "loadFailed": "キャラクターデータの読み込みに失敗しました",
         "renameError": "名前の変更に失敗しました: {{error}}",
diff --git a/static/locales/ko.json b/static/locales/ko.json
@@ -1861,6 +1861,7 @@
         "switchError": "네코를 전환하는 동안 오류가 발생했습니다.",
         "freePresetVoices": "── 무료 사전 설정 음색 ──",
         "geminiNativeVoices": "Gemini 네이티브 음성",
+        "savedVoiceFallback": "현재 저장된 음성",
         "gptsovitsVoices": "GPT-SoVITS 음성",
         "loadFailed": "캐릭터 데이터 로드 실패",
         "renameError": "이름 변경 실패: {{error}}",
diff --git a/static/locales/pt.json b/static/locales/pt.json
@@ -1861,6 +1861,7 @@
         "switchError": "Ocorreu um erro ao trocar de neko",
         "freePresetVoices": "Vozes predefinidas gratuitas",
         "geminiNativeVoices": "Vozes nativas do Gemini",
+        "savedVoiceFallback": "Voz atualmente salva",
         "gptsovitsVoices": "Vozes GPT-SoVITS",
         "loadFailed": "Falha ao carregar dados de caracteres",
         "renameError": "Falha ao renomear: {{error}}",
diff --git a/static/locales/ru.json b/static/locales/ru.json
@@ -1861,6 +1861,7 @@
         "switchError": "Ошибка при переключении NEKO",
         "freePresetVoices": "Бесплатные предустановленные голоса",
         "geminiNativeVoices": "Нативные голоса Gemini",
+        "savedVoiceFallback": "Текущий сохранённый голос",
         "gptsovitsVoices": "Голоса GPT-SoVITS",
         "loadFailed": "Ошибка загрузки данных персонажа",
         "renameError": "Ошибка переименования: {{error}}",
diff --git a/static/locales/zh-CN.json b/static/locales/zh-CN.json
@@ -1861,6 +1861,7 @@
         "switchError": "切换猫娘时发生错误",
         "freePresetVoices": "免费预设音色",
         "geminiNativeVoices": "Gemini 原生音色",
+        "savedVoiceFallback": "当前已保存音色",
         "gptsovitsVoices": "GPT-SoVITS 声音",
         "loadFailed": "加载角色数据失败",
         "renameError": "重命名失败: {{error}}",
diff --git a/static/locales/zh-TW.json b/static/locales/zh-TW.json
@@ -1861,6 +1861,7 @@
         "switchError": "切換貓娘時發生錯誤",
         "freePresetVoices": "── 免費預設音色 ──",
         "geminiNativeVoices": "Gemini 原生音色",
+        "savedVoiceFallback": "目前已保存音色",
         "gptsovitsVoices": "GPT-SoVITS 聲音",
         "loadFailed": "載入角色資料失敗",
         "renameError": "重新命名失敗: {{error}}",
