新增腾讯云 TTS 服务

Dockerfile

@@ -22,10 +22,9 @@ RUN python -m pip install --upgrade pip setuptools wheel && \
 # 激活虚拟环境
 ENV PATH="/opt/venv/bin:$PATH"
 
-# 复制 requirements.txt 并安装 Python 依赖
+# 复制 requirements.txt 并使用镜像安装 Python 依赖
 COPY requirements.txt .
-RUN pip install --no-cache-dir --upgrade pip && \
-    pip install --no-cache-dir -r requirements.txt
+RUN pip install --no-cache-dir -i https://pypi.tuna.tsinghua.edu.cn/simple -r requirements.txt
 
 # 运行阶段
 FROM python:3.12-slim-bookworm
@@ -48,7 +47,7 @@ ENV PATH="/opt/venv/bin:$PATH" \
     LANG=C.UTF-8 \
     LC_ALL=C.UTF-8
 
-# 安装运行时系统依赖
+# 一次性安装所有依赖、创建用户、配置系统，减少层级
 RUN apt-get update && apt-get install -y --no-install-recommends \
     imagemagick \
     ffmpeg \
@@ -56,32 +55,25 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
     curl \
     git-lfs \
     ca-certificates \
+    dos2unix \
+    && sed -i 's/<policy domain="path" rights="none" pattern="@\*"/<policy domain="path" rights="read|write" pattern="@\*"/' /etc/ImageMagick-6/policy.xml || true \
+    && git lfs install \
+    && groupadd -r narratoai && useradd -r -g narratoai -d /NarratoAI -s /bin/bash narratoai \
     && rm -rf /var/lib/apt/lists/*
 
-# 配置 ImageMagick 策略（允许处理更多格式）
-RUN sed -i 's/<policy domain="path" rights="none" pattern="@\*"/<policy domain="path" rights="read|write" pattern="@\*"/' /etc/ImageMagick-6/policy.xml || true
+# 复制入口脚本并修复换行符问题
+COPY --chown=narratoai:narratoai docker-entrypoint.sh /usr/local/bin/
+RUN dos2unix /usr/local/bin/docker-entrypoint.sh && chmod +x /usr/local/bin/docker-entrypoint.sh
 
-# 初始化 git-lfs
-RUN git lfs install
-
-# 创建非 root 用户（安全最佳实践）
-RUN groupadd -r narratoai && useradd -r -g narratoai -d /NarratoAI -s /bin/bash narratoai
-
-# 复制应用代码
+# 复制其余的应用代码
 COPY --chown=narratoai:narratoai . .
 
-# 确保配置文件存在
-RUN if [ ! -f config.toml ]; then cp config.example.toml config.toml; fi
-
-# 创建必要的目录并设置权限
+# 创建目录、复制配置、设置权限
 RUN mkdir -p storage/temp storage/tasks storage/json storage/narration_scripts storage/drama_analysis && \
+    if [ ! -f config.toml ]; then cp config.example.toml config.toml; fi && \
     chown -R narratoai:narratoai /NarratoAI && \
     chmod -R 755 /NarratoAI
 
-# 复制并设置入口点脚本
-COPY --chown=narratoai:narratoai docker-entrypoint.sh /usr/local/bin/
-RUN chmod +x /usr/local/bin/docker-entrypoint.sh
-
 # 切换到非 root 用户
 USER narratoai
 
@@ -93,5 +85,5 @@ HEALTHCHECK --interval=30s --timeout=10s --start-period=60s --retries=3 \
     CMD curl -f http://localhost:8501/_stcore/health || exit 1
 
 # 设置入口点
-ENTRYPOINT ["docker-entrypoint.sh"]
-CMD ["webui"]
+ENTRYPOINT ["/usr/local/bin/docker-entrypoint.sh"]
+CMD ["webui"]

app/config/config.py

@@ -48,6 +48,7 @@ def save_config():
     with open(config_file, "w", encoding="utf-8") as f:
         _cfg["app"] = app
         _cfg["azure"] = azure
+        _cfg["tencent"] = tencent
         _cfg["soulvoice"] = soulvoice
         _cfg["ui"] = ui
         f.write(toml.dumps(_cfg))
@@ -58,6 +59,7 @@ app = _cfg.get("app", {})
 whisper = _cfg.get("whisper", {})
 proxy = _cfg.get("proxy", {})
 azure = _cfg.get("azure", {})
+tencent = _cfg.get("tencent", {})
 soulvoice = _cfg.get("soulvoice", {})
 ui = _cfg.get("ui", {})
 frames = _cfg.get("frames", {})

app/models/schema.py

@@ -176,7 +176,7 @@ class VideoClipParams(BaseModel):
     voice_volume: Optional[float] = Field(default=AudioVolumeDefaults.VOICE_VOLUME, description="解说语音音量")
     voice_rate: Optional[float] = Field(default=1.0, description="语速")
     voice_pitch: Optional[float] = Field(default=1.0, description="语调")
-
+    tts_engine: Optional[str] = Field(default="tencent", description="TTS 引擎")
     bgm_name: Optional[str] = Field(default="random", description="背景音乐名称")
     bgm_type: Optional[str] = Field(default="random", description="背景音乐类型")
     bgm_file: Optional[str] = Field(default="", description="背景音乐文件")

app/services/clip_video.py

@@ -18,7 +18,6 @@ from pathlib import Path
 
 from app.utils import ffmpeg_utils
 
-
 def parse_timestamp(timestamp: str) -> tuple:
     """
     解析时间戳字符串，返回开始和结束时间

app/services/llm/providers/__init__.py

@@ -43,5 +43,5 @@ __all__ = [
     'QwenTextProvider',
     'DeepSeekTextProvider',
     'SiliconflowVisionProvider',
-    'SiliconflowTextProvider'
+    'SiliconflowTextProvider',
 ]

app/services/task.py

@@ -73,6 +73,7 @@ def start_subclip(task_id: str, params: VideoClipParams, subclip_path_videos: di
     tts_results = voice.tts_multiple(
         task_id=task_id,
         list_script=tts_segments,  # 只传入需要TTS的片段
+        tts_engine=params.tts_engine,
         voice_name=params.voice_name,
         voice_rate=params.voice_rate,
         voice_pitch=params.voice_pitch,
@@ -317,6 +318,7 @@ def start_subclip_unified(task_id: str, params: VideoClipParams):
     tts_results = voice.tts_multiple(
         task_id=task_id,
         list_script=tts_segments,  # 只传入需要TTS的片段
+        tts_engine=params.tts_engine,
         voice_name=params.voice_name,
         voice_rate=params.voice_rate,
         voice_pitch=params.voice_pitch,

app/services/voice.py

@@ -5,6 +5,7 @@ import traceback
 import edge_tts
 import asyncio
 import requests
+import uuid
 from loguru import logger
 from typing import List, Union, Tuple
 from datetime import datetime
@@ -1080,17 +1081,27 @@ def should_use_azure_speech_services(voice_name: str) -> bool:
 
 
 def tts(
-    text: str, voice_name: str, voice_rate: float, voice_pitch: float, voice_file: str
+    text: str, voice_name: str, voice_rate: float, voice_pitch: float, voice_file: str, tts_engine: str = "azure"
 ) -> Union[SubMaker, None]:
-    # 检查是否为 SoulVoice 引擎
-    if is_soulvoice_voice(voice_name):
+    logger.info(f"使用 TTS 引擎: '{tts_engine}', 语音: '{voice_name}'")
+
+    if tts_engine == "tencent":
+        logger.info("分发到腾讯云 TTS")
+        return tencent_tts(text, voice_name, voice_file, speed=voice_rate)
+    
+    if tts_engine == "soulvoice":
+        logger.info("分发到 SoulVoice TTS")
         return soulvoice_tts(text, voice_name, voice_file, speed=voice_rate)
 
-    # 检查是否应该使用 Azure Speech Services
-    if should_use_azure_speech_services(voice_name):
-        return azure_tts_v2(text, voice_name, voice_file)
+    if tts_engine == "azure":
+        if should_use_azure_speech_services(voice_name):
+            logger.info("分发到 Azure Speech Services (V2)")
+            return azure_tts_v2(text, voice_name, voice_file)
+        logger.info("分发到 Edge TTS (Azure V1)")
+        return azure_tts_v1(text, voice_name, voice_rate, voice_pitch, voice_file)
 
-    # 默认使用 Edge TTS (Azure V1)
+    # Fallback for unknown engine - default to azure v1
+    logger.warning(f"未知的 TTS 引擎: '{tts_engine}', 将默认使用 Edge TTS (Azure V1)。")
     return azure_tts_v1(text, voice_name, voice_rate, voice_pitch, voice_file)
 
 
@@ -1483,7 +1494,7 @@ def get_audio_duration(sub_maker: submaker.SubMaker):
     return sub_maker.offset[-1][1] / 10000000
 
 
-def tts_multiple(task_id: str, list_script: list, voice_name: str, voice_rate: float, voice_pitch: float):
+def tts_multiple(task_id: str, list_script: list, voice_name: str, voice_rate: float, voice_pitch: float, tts_engine: str = "azure"):
     """
     根据JSON文件中的多段文本进行TTS转换
     
@@ -1491,6 +1502,7 @@ def tts_multiple(task_id: str, list_script: list, voice_name: str, voice_rate: f
     :param list_script: 脚本列表
     :param voice_name: 语音名称
     :param voice_rate: 语音速率
+    :param tts_engine: TTS 引擎
     :return: 生成的音频文件列表
     """
     voice_name = parse_voice_name(voice_name)
@@ -1512,6 +1524,7 @@ def tts_multiple(task_id: str, list_script: list, voice_name: str, voice_rate: f
                 voice_rate=voice_rate,
                 voice_pitch=voice_pitch,
                 voice_file=audio_file,
+                tts_engine=tts_engine,
             )
 
             if sub_maker is None:
@@ -1581,14 +1594,6 @@ def get_audio_duration_from_file(audio_file: str) -> float:
         # 如果所有方法都失败，返回一个基于文本长度的估算
         return 3.0  # 默认3秒，避免返回0
 
-
-def is_soulvoice_voice(voice_name: str) -> bool:
-    """
-    检查是否为 SoulVoice 语音
-    """
-    return voice_name.startswith("soulvoice:") or voice_name.startswith("speech:")
-
-
 def parse_soulvoice_voice(voice_name: str) -> str:
     """
     解析 SoulVoice 语音名称
@@ -1600,6 +1605,118 @@ def parse_soulvoice_voice(voice_name: str) -> str:
         return voice_name[10:]  # 移除 "soulvoice:" 前缀
     return voice_name
 
+def parse_tencent_voice(voice_name: str) -> str:
+    """
+    解析腾讯云 TTS 语音名称
+    支持格式：tencent:101001
+    """
+    if voice_name.startswith("tencent:"):
+        return voice_name[8:]  # 移除 "tencent:" 前缀
+    return voice_name
+
+
+def tencent_tts(text: str, voice_name: str, voice_file: str, speed: float = 1.0) -> Union[SubMaker, None]:
+    """
+    使用腾讯云 TTS 生成语音
+    """
+    try:
+        # 导入腾讯云 SDK
+        from tencentcloud.common import credential
+        from tencentcloud.common.profile.client_profile import ClientProfile
+        from tencentcloud.common.profile.http_profile import HttpProfile
+        from tencentcloud.tts.v20190823 import tts_client, models
+        import base64
+    except ImportError as e:
+        logger.error(f"腾讯云 SDK 未安装: {e}")
+        return None
+    
+    # 获取腾讯云配置
+    tencent_config = config.tencent
+    secret_id = tencent_config.get("secret_id")
+    secret_key = tencent_config.get("secret_key")
+    region = tencent_config.get("region", "ap-beijing")
+    
+    if not secret_id or not secret_key:
+        logger.error("腾讯云 TTS 配置不完整，请检查 secret_id 和 secret_key")
+        return None
+    
+    # 解析语音名称
+    voice_type = parse_tencent_voice(voice_name)
+    
+    # 转换速度参数 (腾讯云支持 -2 到 2 的范围)
+    speed_value = max(-2.0, min(2.0, (speed - 1.0) * 2))
+    
+    for i in range(3):
+        try:
+            logger.info(f"第 {i+1} 次使用腾讯云 TTS 生成音频")
+            
+            # 创建认证对象
+            cred = credential.Credential(secret_id, secret_key)
+            
+            # 创建 HTTP 配置
+            httpProfile = HttpProfile()
+            httpProfile.endpoint = "tts.tencentcloudapi.com"
+            
+            # 创建客户端配置
+            clientProfile = ClientProfile()
+            clientProfile.httpProfile = httpProfile
+            
+            # 创建客户端
+            client = tts_client.TtsClient(cred, region, clientProfile)
+            
+            req = models.TextToVoiceRequest()
+            req.Text = text
+            req.SessionId = str(uuid.uuid4())
+            req.VoiceType = int(voice_type) if voice_type.isdigit() else 101001
+            req.Speed = speed_value
+            req.SampleRate = 16000
+            req.Codec = "mp3"
+            req.ProjectId = 0
+            req.ModelType = 1
+            req.PrimaryLanguage = 1
+            req.EnableSubtitle = True
+
+            # 发送请求
+            resp = client.TextToVoice(req)
+            
+            # 检查响应
+            if not resp.Audio:
+                logger.warning(f"腾讯云 TTS 返回空音频数据")
+                if i < 2:
+                    time.sleep(1)
+                continue
+            
+            # 解码音频数据
+            audio_data = base64.b64decode(resp.Audio)
+            
+            # 写入文件
+            with open(voice_file, "wb") as f:
+                f.write(audio_data)
+
+            # 创建字幕对象
+            sub_maker = SubMaker()
+            if resp.Subtitles:
+                for sub in resp.Subtitles:
+                    start_ms = sub.BeginTime
+                    end_ms = sub.EndTime
+                    text = sub.Text
+                    # 转换为 100ns 单位
+                    sub_maker.create_sub((start_ms * 10000, end_ms * 10000), text)
+            else:
+                # 如果没有字幕返回，则使用估算作为后备方案
+                duration_ms = len(text) * 200
+                sub_maker.create_sub((0, duration_ms * 10000), text)
+
+            logger.info(f"腾讯云 TTS 生成成功，文件大小: {len(audio_data)} 字节")
+            return sub_maker
+
+        except Exception as e:
+            logger.error(f"腾讯云 TTS 生成音频时出错: {str(e)}")
+            if i < 2:
+                 time.sleep(1)
+     
+    return None
+
 
 def soulvoice_tts(text: str, voice_name: str, voice_file: str, speed: float = 1.0) -> Union[SubMaker, None]:
     """

config.example.toml

@@ -96,6 +96,14 @@
     speech_key = ""
     speech_region = ""
 
+[tencent]
+    # 腾讯云 TTS 配置
+    # 访问 https://console.cloud.tencent.com/cam/capi 获取你的密钥
+    secret_id = ""
+    secret_key = ""
+    # 地域配置，默认为 ap-beijing
+    region = "ap-beijing"
+
 [soulvoice]
     # SoulVoice TTS API 密钥
     api_key = ""
@@ -107,7 +115,7 @@
     model = "FunAudioLLM/CosyVoice2-0.5B"
 
 [ui]
-    # TTS引擎选择 (edge_tts, azure_speech, soulvoice)
+    # TTS引擎选择 (edge_tts, azure_speech, soulvoice, tencent_tts)
     tts_engine = "edge_tts"
 
     # Edge TTS 配置

docker-entrypoint.sh

@@ -6,6 +6,61 @@ log() {
     echo "[$(date +'%Y-%m-%d %H:%M:%S')] $1"
 }
 
+# 函数：安装运行时依赖
+install_runtime_dependencies() {
+    log "检查并安装运行时依赖..."
+    
+    # 检查是否需要安装新的依赖
+    local requirements_file="requirements.txt"
+    local installed_packages_file="/tmp/installed_packages.txt"
+    
+    # 如果requirements.txt存在且比已安装包列表新，则重新安装
+    if [ -f "$requirements_file" ]; then
+        if [ ! -f "$installed_packages_file" ] || [ "$requirements_file" -nt "$installed_packages_file" ]; then
+            log "发现新的依赖需求，开始安装..."
+            
+            # 尝试使用sudo安装，如果失败则使用用户级安装
+            if command -v sudo >/dev/null 2>&1 && sudo -n true 2>/dev/null; then
+                log "尝试使用sudo安装依赖..."
+                sudo pip install --no-cache-dir -r "$requirements_file" 2>&1 | while read line; do
+                    log "pip: $line"
+                done
+                INSTALL_RESULT=${PIPESTATUS[0]}
+            else
+                INSTALL_RESULT=1  # 设置为失败，触发用户级安装
+            fi
+            
+            # 如果sudo安装失败，尝试用户级安装
+            if [ $INSTALL_RESULT -ne 0 ]; then
+                log "尝试用户级安装依赖..."
+                pip install --user --no-cache-dir -r "$requirements_file" 2>&1 | while read line; do
+                    log "pip: $line"
+                done
+                
+                # 确保用户级安装的包在PATH中
+                export PATH="$HOME/.local/bin:$PATH"
+            fi
+            
+            # 单独安装腾讯云SDK（确保安装）
+            log "确保腾讯云SDK已安装..."
+            if ! pip list | grep -q "tencentcloud-sdk-python"; then
+                log "安装腾讯云SDK..."
+                pip install --user tencentcloud-sdk-python>=3.0.1200
+            else
+                log "腾讯云SDK已安装"
+            fi
+            
+            # 记录安装时间
+            touch "$installed_packages_file"
+            log "依赖安装完成"
+        else
+            log "依赖已是最新版本，跳过安装"
+        fi
+    else
+        log "未找到 requirements.txt 文件"
+    fi
+}
+
 # 函数：检查必要的文件和目录
 check_requirements() {
     log "检查应用环境..."
@@ -27,6 +82,9 @@ check_requirements() {
             mkdir -p "$dir"
         fi
     done
+    
+    # 安装运行时依赖
+    install_runtime_dependencies
 
     log "环境检查完成"
 }

requirements.txt

@@ -14,6 +14,7 @@ pysrt==1.1.2
 openai>=1.77.0
 google-generativeai>=0.8.5
 azure-cognitiveservices-speech>=1.37.0
+tencentcloud-sdk-python>=3.0.1200
 
 # 图像处理依赖
 Pillow>=10.3.0

webui/components/audio_settings.py

@@ -24,7 +24,8 @@ def get_tts_engine_options():
     return {
         "edge_tts": "Edge TTS",
         "azure_speech": "Azure Speech Services",
-        "soulvoice": "SoulVoice"
+        "soulvoice": "SoulVoice",
+        "tencent_tts": "腾讯云 TTS"
     }
 
 
@@ -48,6 +49,12 @@ def get_tts_engine_descriptions():
             "features": "提供免费额度，支持语音克隆，支持微信购买额度，无需信用卡，性价比极高",
             "use_case": "个人用户和中小企业，需要语音克隆功能",
             "registration": "https://soulvoice.scsmtech.cn/"
+        },
+        "tencent_tts": {
+            "title": "腾讯云 TTS",
+            "features": "提供免费额度，音质优秀，支持多种音色，国内访问速度快",
+            "use_case": "个人和企业用户，需要稳定的中文语音合成",
+            "registration": "https://console.cloud.tencent.com/tts"
         }
     }
 
@@ -126,6 +133,8 @@ def render_tts_settings(tr):
         render_azure_speech_settings(tr)
     elif selected_engine == "soulvoice":
         render_soulvoice_engine_settings(tr)
+    elif selected_engine == "tencent_tts":
+        render_tencent_tts_settings(tr)
 
     # 4. 试听功能
     render_voice_preview_new(tr, selected_engine)
@@ -357,6 +366,117 @@ def render_azure_speech_settings(tr):
         st.warning("⚠️ 请配置 API Key")
 
 
+def render_tencent_tts_settings(tr):
+    """渲染腾讯云 TTS 引擎设置"""
+    # Secret ID 输入
+    secret_id = st.text_input(
+        "Secret ID",
+        value=config.tencent.get("secret_id", ""),
+        help="请输入您的腾讯云 Secret ID"
+    )
+
+    # Secret Key 输入
+    secret_key = st.text_input(
+        "Secret Key",
+        value=config.tencent.get("secret_key", ""),
+        type="password",
+        help="请输入您的腾讯云 Secret Key"
+    )
+
+    # 地域选择
+    region_options = [
+        "ap-beijing",
+        "ap-shanghai",
+        "ap-guangzhou",
+        "ap-chengdu",
+        "ap-nanjing",
+        "ap-singapore",
+        "ap-hongkong"
+    ]
+    
+    saved_region = config.tencent.get("region", "ap-beijing")
+    if saved_region not in region_options:
+        region_options.append(saved_region)
+    
+    region = st.selectbox(
+        "服务地域",
+        options=region_options,
+        index=region_options.index(saved_region),
+        help="选择腾讯云 TTS 服务地域"
+    )
+
+    # 音色选择
+    voice_type_options = {
+        "101001": "智瑜 - 女声（推荐）",
+        "101002": "智聆 - 女声",
+        "101003": "智美 - 女声",
+        "101004": "智云 - 男声",
+        "101005": "智莉 - 女声",
+        "101006": "智言 - 男声",
+        "101007": "智娜 - 女声",
+        "101008": "智琪 - 女声",
+        "101009": "智芸 - 女声",
+        "101010": "智华 - 男声",
+        "101011": "智燕 - 女声",
+        "101012": "智丹 - 女声",
+        "101013": "智辉 - 男声",
+        "101014": "智宁 - 女声",
+        "101015": "智萌 - 女声",
+        "101016": "智甜 - 女声",
+        "101017": "智蓉 - 女声",
+        "101018": "智靖 - 男声"
+    }
+    
+    saved_voice_type = config.ui.get("tencent_voice_type", "101001")
+    if saved_voice_type not in voice_type_options:
+        voice_type_options[saved_voice_type] = f"自定义音色 ({saved_voice_type})"
+    
+    selected_voice_display = st.selectbox(
+        "音色选择",
+        options=list(voice_type_options.values()),
+        index=list(voice_type_options.keys()).index(saved_voice_type),
+        help="选择腾讯云 TTS 音色"
+    )
+    
+    # 获取实际的音色ID
+    voice_type = list(voice_type_options.keys())[
+        list(voice_type_options.values()).index(selected_voice_display)
+    ]
+    
+    # 语速调节
+    voice_rate = st.slider(
+        "语速调节",
+        min_value=0.5,
+        max_value=2.0,
+        value=config.ui.get("tencent_rate", 1.0),
+        step=0.1,
+        help="调节语音速度 (0.5-2.0)"
+    )
+    
+    # 显示音色说明
+    with st.expander("💡 腾讯云 TTS 音色说明", expanded=False):
+        st.write("**女声音色：**")
+        female_voices = [(k, v) for k, v in voice_type_options.items() if "女声" in v]
+        for voice_id, voice_desc in female_voices[:6]:  # 显示前6个
+            st.write(f"• {voice_desc} (ID: {voice_id})")
+        
+        st.write("")
+        st.write("**男声音色：**")
+        male_voices = [(k, v) for k, v in voice_type_options.items() if "男声" in v]
+        for voice_id, voice_desc in male_voices:
+            st.write(f"• {voice_desc} (ID: {voice_id})")
+        
+        st.write("")
+        st.info("💡 更多音色请参考腾讯云官方文档")
+    
+    # 保存配置
+    config.tencent["secret_id"] = secret_id
+    config.tencent["secret_key"] = secret_key
+    config.tencent["region"] = region
+    config.ui["tencent_voice_type"] = voice_type
+    config.ui["tencent_rate"] = voice_rate
+
+
 def render_soulvoice_engine_settings(tr):
     """渲染 SoulVoice 引擎设置"""
     # API Key 输入
@@ -453,6 +573,11 @@ def render_voice_preview_new(tr, selected_engine):
                     voice_name = voice_uri if voice_uri.startswith("soulvoice:") else f"soulvoice:{voice_uri}"
             voice_rate = 1.0  # SoulVoice 使用默认语速
             voice_pitch = 1.0  # SoulVoice 不支持音调调节
+        elif selected_engine == "tencent_tts":
+            voice_type = config.ui.get("tencent_voice_type", "101001")
+            voice_name = f"tencent:{voice_type}"
+            voice_rate = config.ui.get("tencent_rate", 1.0)
+            voice_pitch = 1.0  # 腾讯云 TTS 不支持音调调节
 
         if not voice_name:
             st.error("请先配置语音设置")