cerevox 3.0.0-beta.21 → 3.0.0-beta.22

package/dist/mcp/servers/zerocut.js

@@ -50,7 +50,6 @@ const constants_1 = require("../../utils/constants");
 const videokit_1 = require("../../utils/videokit");
 const promises_1 = require("node:fs/promises");
 const node_path_1 = __importStar(require("node:path"));
-const doubao_voices_full_1 = require("./helper/doubao_voices_full");
 const node_fs_1 = require("node:fs");
 const coze_1 = require("../../utils/coze");
 const mp3_duration_1 = __importDefault(require("mp3-duration"));
@@ -2463,44 +2462,44 @@ server.registerTool('generate-scene-tts', {
             .optional()
             .default(1.0)
             .describe('The volume of the tts.'),
-        emotion: zod_1.z
-            .enum([
-            'storytelling',
-            'neutral',
-            'excited',
-            'coldness',
-            'angry',
-            'sad',
-            'happy',
-            'surprised',
-            'fear',
-            'depressed',
-            'lovey-dovey',
-            'shy',
-            'comfort',
-            'tension',
-            'tender',
-            'magnetic',
-            'vocal - fry',
-            'ASMR',
-        ])
-            .optional(),
         voiceID: zod_1.z
             .string()
             .describe(`适合作为视频配音的音色ID，除非用户指定，否则你必须已通过 search_voice 工具检查确定该音色确实是存在的。`),
+        context_texts: zod_1.z
+            .array(zod_1.z.string())
+            .default([])
+            .describe(`语音合成的辅助信息，用于模型对话式合成，能更好的体现语音情感
+
+可以探索，比如常见示例有以下几种：
+
+1. 语速调整
+  - context_texts: ["你可以说慢一点吗？"]
+2. 情绪/语气调整
+  - context_texts=["你可以用特别特别痛心的语气说话吗?"]
+  - context_texts=["嗯，你的语气再欢乐一点"]
+3. 音量调整
+  - context_texts=["你嗓门再小点。"]
+4. 音感调整
+  - context_texts=["你能用骄傲的语气来说话吗？"]
+`),
         explicit_language: zod_1.z.enum(['zh', 'en', 'ja']).optional().default('zh'),
     },
-}, async ({ text, sceneIndex, storyBoardFile, skipConsistencyCheck, voiceID, saveToFileName, speed, pitch, volume, emotion, explicit_language, }) => {
+}, async ({ text, sceneIndex, storyBoardFile, skipConsistencyCheck, voiceID, saveToFileName, speed, pitch, volume, context_texts, explicit_language, }) => {
     try {
         // 验证session状态
         const currentSession = await validateSession('generate-scene-tts');
         const validatedFileName = validateFileName(saveToFileName);
         const finalSpeed = speed ?? 1;
         volume = volume ?? 1;
+        const ai = currentSession.ai;
         let scene = null;
         // 校验 text 与 storyboard.json 中场景设定的一致性
         if (sceneIndex && !skipConsistencyCheck) {
             try {
+                const voice = (await ai.listVoices()).find(v => v.id === voiceID);
+                if (!voice) {
+                    return createErrorResponse(`Voice ${voiceID} not found in voice-list. Use search-voices tool to find available voices. 若用户坚持要使用该音色，需跳过一致性检查。`, 'generate-scene-tts');
+                }
                 const storyBoardPath = (0, node_path_1.resolve)(process.env.ZEROCUT_PROJECT_CWD || process.cwd(), projectLocalDir, storyBoardFile);
                 if ((0, node_fs_1.existsSync)(storyBoardPath)) {
                     const storyBoardContent = await (0, promises_1.readFile)(storyBoardPath, 'utf8');
@@ -2551,7 +2550,6 @@ server.registerTool('generate-scene-tts', {
             }
         }
         console.log(`Generating TTS with voice: ${voiceID}, speed: ${finalSpeed}, text: ${text.substring(0, 100)}...`);
-        const ai = currentSession.ai;
         if (voiceID.startsWith('BV0')) {
             throw new Error(`BV0* 系列音色已弃用，你必须已通过 search_voice 工具检查确定该音色确实是存在的。`);
         }
@@ -2563,30 +2561,80 @@ server.registerTool('generate-scene-tts', {
             ? 'volcano'
             : 'minimax';
         let res;
+        let emotion = 'auto';
         if (type === 'volcano') {
-            const voice = doubao_voices_full_1.doubaoVoicesFull.find(v => v.voiceID === voiceID);
-            if (!voice) {
-                return createErrorResponse(`Voice ${voiceID} not found in Doubao voices. Use search-voices tool to find available voices.`, 'generate-scene-tts');
-            }
-            const emotions = voice.emotions || [];
-            if (emotion && !emotions.includes(emotion)) {
-                emotion = 'neutral';
-            }
-            // 修复可能的 emotion 错误情况
-            emotion = emotion || 'neutral';
             volume = Math.max(Math.min(volume, 2.0), 0.5);
             res = await ai.textToSpeechVolc({
                 text: text.trim(),
                 speaker: voiceID,
                 speed: Math.floor(100 * (finalSpeed - 1)),
                 volume: Math.floor(100 * (volume - 1)),
-                emotion,
+                context_texts,
                 explicit_language,
                 voice_to_caption: explicit_language === 'zh' || explicit_language === 'en',
             });
         }
         else {
-            emotion = emotion || 'neutral';
+            emotion = 'neutral';
+            if (context_texts.length > 0) {
+                const prompt = `根据用户输入语音内容和上下文判断语音的情感，选择以下情感**之一**：
+
+"happy", "sad", "angry", "fearful", "disgusted", "surprised", "calm", "fluent", "whisper"
+
+## 要求
+输出 JSON 格式，包含一个 emotion 字段，值为以上情感之一。
+`;
+                const schema = {
+                    name: 'emotion_schema',
+                    schema: {
+                        type: 'object',
+                        properties: {
+                            emotion: {
+                                type: 'string',
+                                enum: [
+                                    'neutral',
+                                    'happy',
+                                    'sad',
+                                    'angry',
+                                    'fearful',
+                                    'disgusted',
+                                    'surprised',
+                                    'calm',
+                                    'fluent',
+                                    'whisper',
+                                ],
+                                description: '用户输入语音的情感',
+                            },
+                        },
+                        required: ['emotion'],
+                    },
+                };
+                const payload = {
+                    model: 'Doubao-Seed-1.6-flash',
+                    messages: [
+                        {
+                            role: 'system',
+                            content: prompt,
+                        },
+                        {
+                            role: 'user',
+                            content: `## 语音内容：
+${text.trim()}
+
+## 语音上下文
+${context_texts.join('\n')}
+`,
+                        },
+                    ],
+                    response_format: {
+                        type: 'json_schema',
+                        json_schema: schema,
+                    },
+                };
+                const completion = await ai.getCompletions(payload);
+                const emotionObj = JSON.parse(completion.choices[0]?.message?.content ?? '{}');
+                emotion = emotionObj.emotion ?? 'neutral';
+            }
             res = await ai.textToSpeech({
                 text: text.trim(),
                 voiceName: voiceID,
@@ -2627,6 +2675,8 @@ server.registerTool('generate-scene-tts', {
                 uri,
                 durationMs: Math.floor((duration || 0) * 1000),
                 text,
+                emotion,
+                context_texts,
                 voiceName: voiceID,
                 speed: finalSpeed,
                 timestamp: new Date().toISOString(),
@@ -2891,91 +2941,49 @@ server.registerTool('get-schema', {
 });
 server.registerTool('search-voices', {
     title: 'Search Voices',
-    description: 'Search voices from doubao_voices_full based on scenes, emotions, languages, and gender.',
+    description: 'Search voices from doubao_voices_full based on languages, and gender. 搜索并选择符合要求的语音，在合适的情况下，优先采用 volcano_tts_2 类型的语音',
     inputSchema: {
-        scenes: zod_1.z
-            .array(zod_1.z.enum([
-            'asmr',
-            'audiobook',
-            'customer_service',
-            'dialect_fun',
-            'dialogue',
-            'kids_content',
-            'news_explainer',
-            'podcast_voiceover',
-            'product_ad',
-            'promo_trailer',
-            'roleplay_drama',
-            'story_narration',
-            'storytelling',
-            'tutorial',
-        ]))
-            .optional()
-            .describe('Filter by scenes (e.g., ["product_ad", "tutorial"]). If not provided, no scene filtering is applied.'),
-        emotions: zod_1.z
+        languages: zod_1.z
             .array(zod_1.z.enum([
-            'ASMR',
-            'affectionate',
-            'angry',
-            'authoritative',
-            'chat',
-            'coldness',
-            'depressed',
-            'excited',
-            'fear',
-            'happy',
-            'hate',
-            'neutral',
-            'sad',
-            'surprised',
-            'warm',
+            'zh',
+            'en',
+            'ja',
+            'ko',
+            'es',
+            'pt',
+            'nl',
+            'vi',
+            'ru',
+            'id',
+            'de',
+            'fr',
+            'it',
+            'ar',
+            'tr',
+            'uk',
         ]))
             .optional()
-            .describe('Filter by emotions (e.g., ["happy", "neutral"]). If not provided, no emotion filtering is applied.'),
-        languages: zod_1.z
-            .array(zod_1.z.string())
-            .optional()
-            .describe('Filter by languages (e.g., ["zh", "en-US"]). If not provided, no language filtering is applied.'),
+            .describe('Filter by languages (e.g., ["zh", "en"]). If not provided, no language filtering is applied.'),
         gender: zod_1.z
             .enum(['male', 'female'])
             .optional()
             .describe('Filter by gender (male or female). If not provided, no gender filtering is applied.'),
     },
-}, async ({ scenes, emotions, languages, gender }) => {
+}, async ({ languages, gender }) => {
     try {
-        let filteredVoices = [...doubao_voices_full_1.doubaoVoicesFull];
-        // Filter by scenes
-        if (scenes && scenes.length > 0) {
-            filteredVoices = filteredVoices.filter(voice => voice.scenes &&
-                voice.scenes.some(scene => scenes.includes(scene)));
-        }
-        // Filter by emotions
-        if (emotions && emotions.length > 0) {
-            filteredVoices = filteredVoices.filter(voice => {
-                // If emotions includes 'neutral', also include voices without emotions field
-                if (emotions.includes('neutral') && !voice.emotions) {
-                    return true;
-                }
-                return (voice.emotions &&
-                    voice.emotions.some(emotion => emotions.includes(emotion)));
-            });
-        }
+        // 验证session状态
+        const currentSession = await validateSession('search-voices');
+        const ai = currentSession.ai;
+        let filteredVoices = await ai.listVoices();
         // Filter by languages
         if (languages && languages.length > 0) {
             filteredVoices = filteredVoices.filter(voice => voice.languages &&
-                voice.languages.some(lang => languages.includes(lang)));
+                voice.languages.some((lang) => languages.includes(lang)));
         }
         // Filter by gender
         if (gender) {
             filteredVoices = filteredVoices.filter(voice => {
-                const voiceId = voice.voiceID.toLowerCase();
-                if (gender === 'male') {
-                    return voiceId.includes('_male_');
-                }
-                else if (gender === 'female') {
-                    return voiceId.includes('_female_');
-                }
-                return true;
+                return voice.gender === gender;
             });
         }
         return {
@@ -2988,8 +2996,6 @@ server.registerTool('search-voices', {
                             totalCount: filteredVoices.length,
                             voices: filteredVoices,
                             filters: {
-                                scenes: scenes || null,
-                                emotions: emotions || null,
                                 languages: languages || null,
                                 gender: gender || null,
                             },