commit f7c1bed80e8c79b95a56925ea962ce7ecdc9e4ce
Author: hubian <908234780@qq.com>
Date:   Wed Apr 22 16:21:01 2026 +0800

    feat: ChatTTS对话式语音合成服务初始版本

diff --git a/README.md b/README.md
new file mode 100644
index 0000000..b878d46
--- /dev/null
+++ b/README.md
@@ -0,0 +1,147 @@
+# ChatTTS 语音合成服务
+
+专为对话场景设计的 TTS 模型服务，支持情感控制。
+
+## 端口
+
+- **服务端口**: 12002
+
+## 功能
+
+- 对话式语音合成（有语气变化）
+- 情感控制（开心、悲伤、笑声）
+- 批量合成
+- 中文效果好
+
+## 硬件要求
+
+- **GPU**: 推荐 NVIDIA GPU，显存 ≥ 4GB
+- **CPU**: 可运行但较慢
+- **内存**: ≥ 16GB
+
+## 部署步骤
+
+### 1. 安装依赖
+
+```bash
+pip install -r requirements.txt
+```
+
+### 2. 启动服务
+
+```bash
+# 默认端口 12002
+python3 server.py
+
+# 或使用脚本
+PORT=12002 ./start.sh
+```
+
+首次启动会自动下载 ChatTTS 模型（约 2GB）。
+
+### 3. 验证服务
+
+```bash
+curl http://localhost:12002/
+```
+
+返回：
+```json
+{
+  "status": "ok",
+  "service": "ChatTTS",
+  "model_loaded": true,
+  "device": "cuda"
+}
+```
+
+## API 接口
+
+### 合成语音
+
+```bash
+POST /synthesize
+Content-Type: multipart/form-data
+
+参数:
+- text: 要合成的文本
+- temperature: 温度参数（默认 0.3）
+- top_p: Top-P 参数（默认 0.7）
+- top_k: Top-K 参数（默认 20）
+
+返回:
+{
+  "audio_url": "/audio/xxx.wav",
+  "duration": 3.5,
+  "text": "你好",
+  "timestamp": "..."
+}
+```
+
+**示例**：
+```bash
+curl -X POST http://localhost:12002/synthesize \
+  -F "text=你好，很高兴见到你"
+```
+
+### 带情感合成
+
+```bash
+POST /synthesize/emotion
+
+参数:
+- text: 文本
+- emotion: 情感类型（neutral/happy/sad/laugh）
+
+返回: 同上
+```
+
+**情感类型**：
+- `neutral`: 平静
+- `happy`: 开心
+- `sad`: 悲伤
+- `laugh`: 笑声
+
+### 批量合成
+
+```bash
+POST /synthesize/batch
+Content-Type: application/json
+
+Body:
+[
+  {"text": "第一句"},
+  {"text": "第二句"}
+]
+```
+
+### 获取音频文件
+
+```bash
+GET /audio/{filename}
+```
+
+## 特点对比
+
+| 特点 | Edge TTS | ChatTTS |
+|------|----------|---------|
+| 部署 | 无需部署 | 需 GPU |
+| 质量 | 高 | 对话优化 |
+| 情感 | 固定音色 | 可控 |
+| 延迟 | 1-3秒 | 0.5-1秒 |
+| 离线 | ❌ | ✅ |
+
+## 与语音对话网页集成
+
+在 voice-chat-web 的 TTS 设置中选择 ChatTTS，服务地址配置为：
+
+```
+http://192.168.2.5:12002
+```
+
+## 注意事项
+
+1. 模型首次加载需要下载约 2GB
+2. GPU 运行速度更快，CPU 可用但较慢
+3. 长文本建议分段合成
+4. 情感标记功能需要 ChatTTS 版本支持
\ No newline at end of file
diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 0000000..eaecb69
--- /dev/null
+++ b/requirements.txt
@@ -0,0 +1,7 @@
+fastapi==0.110.0
+uvicorn==0.27.1
+python-multipart==0.0.9
+torch==2.2.0
+torchaudio==2.2.0
+transformers==4.38.0
+ChatTTS==0.1.1
\ No newline at end of file
diff --git a/server.py b/server.py
new file mode 100644
index 0000000..6d71e87
--- /dev/null
+++ b/server.py
@@ -0,0 +1,315 @@
+"""
+ChatTTS 语音合成服务
+专为对话场景设计的 TTS 模型
+"""
+
+import os
+import io
+import uuid
+import logging
+import tempfile
+import wave
+import numpy as np
+from typing import Optional
+from datetime import datetime
+
+import torch
+import torchaudio
+from fastapi import FastAPI, UploadFile, File, HTTPException, Form
+from fastapi.middleware.cors import CORSMiddleware
+from fastapi.responses import FileResponse
+from pydantic import BaseModel
+
+# 配置
+PORT = int(os.getenv("PORT", "12002"))
+SAMPLE_RATE = 24000  # ChatTTS 默认采样率
+AUDIO_DIR = os.getenv("AUDIO_DIR", "audio_output")
+os.makedirs(AUDIO_DIR, exist_ok=True)
+
+# 日志
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+app = FastAPI(
+    title="ChatTTS Service",
+    description="对话式语音合成服务",
+    version="1.0.0"
+)
+
+# CORS
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+# ChatTTS 模型（延迟加载）
+chat_model = None
+
+
+class SynthesizeRequest(BaseModel):
+    """合成请求"""
+    text: str
+    temperature: float = 0.3
+    top_p: float = 0.7
+    top_k: int = 20
+
+
+class SynthesizeResponse(BaseModel):
+    """合成响应"""
+    audio_url: str
+    duration: float
+    text: str
+    timestamp: str
+
+
+def load_model():
+    """加载 ChatTTS 模型"""
+    global chat_model
+    if chat_model is None:
+        logger.info("Loading ChatTTS model...")
+        try:
+            import ChatTTS
+            chat_model = ChatTTS.Chat()
+            
+            # 加载模型
+            # 方式一：从本地加载（如果有）
+            # 方式二：自动下载
+            chat_model.load(
+                compile=True,  # 编译优化
+                device="cuda" if torch.cuda.is_available() else "cpu"
+            )
+            
+            logger.info("ChatTTS model loaded successfully")
+        except Exception as e:
+            logger.error(f"Failed to load ChatTTS: {e}")
+            raise
+    return chat_model
+
+
+def save_audio(audio_tensor: torch.Tensor, filename: str) -> str:
+    """
+    保存音频文件
+    audio_tensor: shape [1, samples] 或 [samples]
+    """
+    filepath = os.path.join(AUDIO_DIR, filename)
+    
+    # 确保 tensor 正确形状
+    if audio_tensor.dim() == 1:
+        audio_tensor = audio_tensor.unsqueeze(0)
+    
+    # 保存为 WAV
+    torchaudio.save(filepath, audio_tensor, SAMPLE_RATE, format="wav")
+    
+    return filepath
+
+
+@app.on_event("startup")
+async def startup():
+    """启动时预加载模型"""
+    try:
+        load_model()
+        logger.info(f"ChatTTS service ready on port {PORT}")
+    except Exception as e:
+        logger.warning(f"Model load delayed: {e}")
+
+
+@app.get("/")
+async def root():
+    """健康检查"""
+    return {
+        "status": "ok",
+        "service": "ChatTTS",
+        "model_loaded": chat_model is not None,
+        "device": "cuda" if torch.cuda.is_available() else "cpu"
+    }
+
+
+@app.get("/health")
+async def health():
+    """健康检查"""
+    model_loaded = chat_model is not None
+    return {
+        "status": "ok" if model_loaded else "loading",
+        "gpu": torch.cuda.is_available()
+    }
+
+
+@app.post("/synthesize", response_model=SynthesizeResponse)
+async def synthesize(
+    text: str = Form(..., description="要合成的文本"),
+    temperature: float = Form(0.3, description="温度参数"),
+    top_p: float = Form(0.7, description="Top-P参数"),
+    top_k: int = Form(20, description="Top-K参数")
+):
+    """
+    合成语音
+    
+    ChatTTS 特点：
+    - 支持对话场景，有语气变化
+    - 支持笑声、叹气等情感
+    - 中文效果好
+    """
+    try:
+        model = load_model()
+        
+        # 生成唯一文件名
+        filename = f"{uuid.uuid4().hex}.wav"
+        
+        # 合成参数
+        params = {
+            'temperature': temperature,
+            'top_p': top_p,
+            'top_k': top_k,
+            'spk_emb': None,  # 可选：说话人嵌入
+        }
+        
+        # 合成语音
+        logger.info(f"Synthesizing: {text[:50]}...")
+        
+        # ChatTTS 生成
+        audio_tensor = model.infer(
+            [text],
+            params=params
+        )[0]  # 返回是列表，取第一个
+        
+        # 保存音频
+        filepath = save_audio(audio_tensor, filename)
+        
+        # 计算时长
+        duration = audio_tensor.shape[-1] / SAMPLE_RATE
+        
+        audio_url = f"/audio/{filename}"
+        
+        logger.info(f"Generated audio: {duration:.2f}s")
+        
+        return SynthesizeResponse(
+            audio_url=audio_url,
+            duration=round(duration, 2),
+            text=text,
+            timestamp=datetime.now().isoformat()
+        )
+        
+    except Exception as e:
+        logger.error(f"Synthesis error: {e}", exc_info=True)
+        raise HTTPException(status_code=500, detail=str(e))
+
+
+@app.post("/synthesize/batch")
+async def synthesize_batch(requests: list[SynthesizeRequest]):
+    """
+    批量合成语音
+    """
+    try:
+        model = load_model()
+        
+        texts = [r.text for r in requests]
+        
+        # 统一参数
+        params = {
+            'temperature': requests[0].temperature,
+            'top_p': requests[0].top_p,
+            'top_k': requests[0].top_k,
+        }
+        
+        # 批量生成
+        audio_tensors = model.infer(texts, params=params)
+        
+        results = []
+        for i, audio_tensor in enumerate(audio_tensors):
+            filename = f"{uuid.uuid4().hex}.wav"
+            filepath = save_audio(audio_tensor, filename)
+            duration = audio_tensor.shape[-1] / SAMPLE_RATE
+            
+            results.append({
+                "audio_url": f"/audio/{filename}",
+                "duration": round(duration, 2),
+                "text": texts[i]
+            })
+        
+        return {"results": results}
+        
+    except Exception as e:
+        logger.error(f"Batch synthesis error: {e}")
+        raise HTTPException(status_code=500, detail=str(e))
+
+
+@app.get("/audio/{filename}")
+async def get_audio(filename: str):
+    """获取音频文件"""
+    filepath = os.path.join(AUDIO_DIR, filename)
+    if os.path.exists(filepath):
+        return FileResponse(filepath, media_type="audio/wav")
+    raise HTTPException(status_code=404, detail="Audio not found")
+
+
+@app.delete("/audio/{filename}")
+async def delete_audio(filename: str):
+    """删除音频文件"""
+    filepath = os.path.join(AUDIO_DIR, filename)
+    if os.path.exists(filepath):
+        os.remove(filepath)
+        return {"status": "deleted"}
+    return {"status": "not_found"}
+
+
+# 情感控制接口（ChatTTS 特色）
+@app.post("/synthesize/emotion")
+async def synthesize_with_emotion(
+    text: str = Form(...),
+    emotion: str = Form("neutral", description="情感：neutral/happy/sad/laugh")
+):
+    """
+    带情感的语音合成
+    
+    ChatTTS 支持情感控制：
+    - neutral: 平静
+    - happy: 开心
+    - sad: 悲伤
+    - laugh: 笑声
+    """
+    try:
+        model = load_model()
+        
+        # 根据情感调整参数
+        emotion_params = {
+            'neutral': {'temperature': 0.3, 'top_p': 0.7},
+            'happy': {'temperature': 0.5, 'top_p': 0.8},
+            'sad': {'temperature': 0.2, 'top_p': 0.6},
+            'laugh': {'temperature': 0.6, 'top_p': 0.9},
+        }
+        
+        params = emotion_params.get(emotion, emotion_params['neutral'])
+        
+        # 添加情感标记（ChatTTS 特有）
+        if emotion == 'laugh':
+            # 添加笑声标记
+            text = f"[laugh]{text}"
+        elif emotion == 'happy':
+            text = f"[happy]{text}"
+        elif emotion == 'sad':
+            text = f"[sad]{text}"
+        
+        filename = f"{uuid.uuid4().hex}.wav"
+        audio_tensor = model.infer([text], params=params)[0]
+        
+        filepath = save_audio(audio_tensor, filename)
+        duration = audio_tensor.shape[-1] / SAMPLE_RATE
+        
+        return SynthesizeResponse(
+            audio_url=f"/audio/{filename}",
+            duration=round(duration, 2),
+            text=text,
+            timestamp=datetime.now().isoformat()
+        )
+        
+    except Exception as e:
+        logger.error(f"Emotion synthesis error: {e}")
+        raise HTTPException(status_code=500, detail=str(e))
+
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=PORT)
\ No newline at end of file
diff --git a/start.sh b/start.sh
new file mode 100755
index 0000000..8d2850f
--- /dev/null
+++ b/start.sh
@@ -0,0 +1,11 @@
+#!/bin/bash
+# ChatTTS 服务启动脚本
+
+cd "$(dirname "$0")"
+
+PORT=${PORT:-12002}
+
+echo "Starting ChatTTS service on port $PORT..."
+echo "Device: $(python3 -c 'import torch; print("cuda" if torch.cuda.is_available() else "cpu")')"
+
+python3 server.py
\ No newline at end of file