feat: increase siliconflow TTS services

app/config/config.py

@@ -36,6 +36,7 @@ def save_config():
     with open(config_file, "w", encoding="utf-8") as f:
         _cfg["app"] = app
         _cfg["azure"] = azure
+        _cfg["siliconflow"] = siliconflow
         _cfg["ui"] = ui
         f.write(toml.dumps(_cfg))
 
@@ -45,9 +46,13 @@ app = _cfg.get("app", {})
 whisper = _cfg.get("whisper", {})
 proxy = _cfg.get("proxy", {})
 azure = _cfg.get("azure", {})
-ui = _cfg.get("ui", {
-    "hide_log": False,
-})
+siliconflow = _cfg.get("siliconflow", {})
+ui = _cfg.get(
+    "ui",
+    {
+        "hide_log": False,
+    },
+)
 
 hostname = socket.gethostname()
 

app/services/voice.py

@@ -6,6 +6,7 @@ from typing import Union
 from xml.sax.saxutils import unescape
 
 import edge_tts
+import requests
 from edge_tts import SubMaker, submaker
 from edge_tts.submaker import mktimestamp
 from loguru import logger
@@ -15,6 +16,32 @@ from app.config import config
 from app.utils import utils
 
 
+def get_siliconflow_voices() -> list[str]:
+    """
+    获取硅基流动的声音列表
+
+    Returns:
+        声音列表，格式为 ["siliconflow:FunAudioLLM/CosyVoice2-0.5B:alex", ...]
+    """
+    # 硅基流动的声音列表和对应的性别（用于显示）
+    voices_with_gender = [
+        ("FunAudioLLM/CosyVoice2-0.5B", "alex", "Male"),
+        ("FunAudioLLM/CosyVoice2-0.5B", "anna", "Female"),
+        ("FunAudioLLM/CosyVoice2-0.5B", "bella", "Female"),
+        ("FunAudioLLM/CosyVoice2-0.5B", "benjamin", "Male"),
+        ("FunAudioLLM/CosyVoice2-0.5B", "charles", "Male"),
+        ("FunAudioLLM/CosyVoice2-0.5B", "claire", "Female"),
+        ("FunAudioLLM/CosyVoice2-0.5B", "david", "Male"),
+        ("FunAudioLLM/CosyVoice2-0.5B", "diana", "Female"),
+    ]
+
+    # 添加siliconflow:前缀，并格式化为显示名称
+    return [
+        f"siliconflow:{model}:{voice}-{gender}"
+        for model, voice, gender in voices_with_gender
+    ]
+
+
 def get_all_azure_voices(filter_locals=None) -> list[str]:
     azure_voices_str = """
 Name: af-ZA-AdriNeural
@@ -1045,11 +1072,37 @@ def is_azure_v2_voice(voice_name: str):
     return ""
 
 
+def is_siliconflow_voice(voice_name: str):
+    """检查是否是硅基流动的声音"""
+    return voice_name.startswith("siliconflow:")
+
+
 def tts(
-    text: str, voice_name: str, voice_rate: float, voice_file: str
+    text: str,
+    voice_name: str,
+    voice_rate: float,
+    voice_file: str,
+    voice_volume: float = 1.0,
 ) -> Union[SubMaker, None]:
     if is_azure_v2_voice(voice_name):
         return azure_tts_v2(text, voice_name, voice_file)
+    elif is_siliconflow_voice(voice_name):
+        # 从voice_name中提取模型和声音
+        # 格式: siliconflow:model:voice-Gender
+        parts = voice_name.split(":")
+        if len(parts) >= 3:
+            model = parts[1]
+            # 移除性别后缀，例如 "alex-Male" -> "alex"
+            voice_with_gender = parts[2]
+            voice = voice_with_gender.split("-")[0]
+            # 构建完整的voice参数，格式为 "model:voice"
+            full_voice = f"{model}:{voice}"
+            return siliconflow_tts(
+                text, model, full_voice, voice_rate, voice_file, voice_volume
+            )
+        else:
+            logger.error(f"Invalid siliconflow voice name format: {voice_name}")
+            return None
     return azure_tts_v1(text, voice_name, voice_rate, voice_file)
 
 
@@ -1098,6 +1151,144 @@ def azure_tts_v1(
     return None
 
 
+def siliconflow_tts(
+    text: str,
+    model: str,
+    voice: str,
+    voice_rate: float,
+    voice_file: str,
+    voice_volume: float = 1.0,
+) -> Union[SubMaker, None]:
+    """
+    使用硅基流动的API生成语音
+
+    Args:
+        text: 要转换为语音的文本
+        model: 模型名称，如 "FunAudioLLM/CosyVoice2-0.5B"
+        voice: 声音名称，如 "FunAudioLLM/CosyVoice2-0.5B:alex"
+        voice_rate: 语音速度，范围[0.25, 4.0]
+        voice_file: 输出的音频文件路径
+        voice_volume: 语音音量，范围[0.6, 5.0]，需要转换为硅基流动的增益范围[-10, 10]
+
+    Returns:
+        SubMaker对象或None
+    """
+    text = text.strip()
+    api_key = config.siliconflow.get("api_key", "")
+
+    if not api_key:
+        logger.error("SiliconFlow API key is not set")
+        return None
+
+    # 将voice_volume转换为硅基流动的增益范围
+    # 默认voice_volume为1.0，对应gain为0
+    gain = voice_volume - 1.0
+    # 确保gain在[-10, 10]范围内
+    gain = max(-10, min(10, gain))
+
+    url = "https://api.siliconflow.cn/v1/audio/speech"
+
+    payload = {
+        "model": model,
+        "input": text,
+        "voice": voice,
+        "response_format": "mp3",
+        "sample_rate": 32000,
+        "stream": False,
+        "speed": voice_rate,
+        "gain": gain,
+    }
+
+    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
+
+    for i in range(3):  # 尝试3次
+        try:
+            logger.info(
+                f"start siliconflow tts, model: {model}, voice: {voice}, try: {i + 1}"
+            )
+
+            response = requests.post(url, json=payload, headers=headers)
+
+            if response.status_code == 200:
+                # 保存音频文件
+                with open(voice_file, "wb") as f:
+                    f.write(response.content)
+
+                # 创建一个空的SubMaker对象
+                sub_maker = SubMaker()
+
+                # 获取音频文件的实际长度
+                try:
+                    # 尝试使用moviepy获取音频长度
+                    from moviepy import AudioFileClip
+
+                    audio_clip = AudioFileClip(voice_file)
+                    audio_duration = audio_clip.duration
+                    audio_clip.close()
+
+                    # 将音频长度转换为100纳秒单位（与edge_tts兼容）
+                    audio_duration_100ns = int(audio_duration * 10000000)
+
+                    # 使用文本分割来创建更准确的字幕
+                    # 将文本按标点符号分割成句子
+                    sentences = utils.split_string_by_punctuations(text)
+
+                    if sentences:
+                        # 计算每个句子的大致时长（按字符数比例分配）
+                        total_chars = sum(len(s) for s in sentences)
+                        char_duration = (
+                            audio_duration_100ns / total_chars if total_chars > 0 else 0
+                        )
+
+                        current_offset = 0
+                        for sentence in sentences:
+                            if not sentence.strip():
+                                continue
+
+                            # 计算当前句子的时长
+                            sentence_chars = len(sentence)
+                            sentence_duration = int(sentence_chars * char_duration)
+
+                            # 添加到SubMaker
+                            sub_maker.subs.append(sentence)
+                            sub_maker.offset.append(
+                                (current_offset, current_offset + sentence_duration)
+                            )
+
+                            # 更新偏移量
+                            current_offset += sentence_duration
+                    else:
+                        # 如果无法分割，则使用整个文本作为一个字幕
+                        sub_maker.subs = [text]
+                        sub_maker.offset = [(0, audio_duration_100ns)]
+
+                except Exception as e:
+                    logger.warning(f"Failed to create accurate subtitles: {str(e)}")
+                    # 回退到简单的字幕
+                    sub_maker.subs = [text]
+                    # 使用音频文件的实际长度，如果无法获取，则假设为10秒
+                    sub_maker.offset = [
+                        (
+                            0,
+                            audio_duration_100ns
+                            if "audio_duration_100ns" in locals()
+                            else 10000000,
+                        )
+                    ]
+
+                logger.success(f"siliconflow tts succeeded: {voice_file}")
+                print("s", sub_maker.subs, sub_maker.offset)
+                return sub_maker
+            else:
+                logger.error(
+                    f"siliconflow tts failed with status code {response.status_code}: {response.text}"
+                )
+        except Exception as e:
+            logger.error(f"siliconflow tts failed: {str(e)}")
+
+    return None
+
+
 def azure_tts_v2(text: str, voice_name: str, voice_file: str) -> Union[SubMaker, None]:
     voice_name = is_azure_v2_voice(voice_name)
     if not voice_name:

config.example.toml

@@ -193,6 +193,11 @@ compute_type = "int8"
 speech_key = ""
 speech_region = ""
 
+[siliconflow]
+# SiliconFlow API Key
+# Get your API key at https://siliconflow.cn
+api_key = ""
+
 [ui]
 # UI related settings
 # 是否隐藏日志信息

webui/Main.py

@@ -621,6 +621,7 @@ with middle_panel:
         tts_servers = [
             ("azure-tts-v1", "Azure TTS V1"),
             ("azure-tts-v2", "Azure TTS V2"),
+            ("siliconflow", "SiliconFlow TTS"),
         ]
 
         # 获取保存的TTS服务器，默认为v1
@@ -641,20 +642,26 @@ with middle_panel:
         selected_tts_server = tts_servers[selected_tts_server_index][0]
         config.ui["tts_server"] = selected_tts_server
 
-        # 获取所有声音
-        all_voices = voice.get_all_azure_voices(filter_locals=None)
-
-        # 根据选择的TTS服务器筛选声音
+        # 根据选择的TTS服务器获取声音列表
         filtered_voices = []
-        for v in all_voices:
-            if selected_tts_server == "azure-tts-v2":
-                # V2版本的声音名称中包含"v2"
-                if "V2" in v:
-                    filtered_voices.append(v)
-            else:
-                # V1版本的声音名称中不包含"v2"
-                if "V2" not in v:
-                    filtered_voices.append(v)
+
+        if selected_tts_server == "siliconflow":
+            # 获取硅基流动的声音列表
+            filtered_voices = voice.get_siliconflow_voices()
+        else:
+            # 获取Azure的声音列表
+            all_voices = voice.get_all_azure_voices(filter_locals=None)
+
+            # 根据选择的TTS服务器筛选声音
+            for v in all_voices:
+                if selected_tts_server == "azure-tts-v2":
+                    # V2版本的声音名称中包含"v2"
+                    if "V2" in v:
+                        filtered_voices.append(v)
+                else:
+                    # V1版本的声音名称中不包含"v2"
+                    if "V2" not in v:
+                        filtered_voices.append(v)
 
         friendly_names = {
             v: v.replace("Female", tr("Female"))
@@ -720,6 +727,7 @@ with middle_panel:
                     voice_name=voice_name,
                     voice_rate=params.voice_rate,
                     voice_file=audio_file,
+                    voice_volume=params.voice_volume,
                 )
                 # if the voice file generation failed, try again with a default content.
                 if not sub_maker:
@@ -729,6 +737,7 @@ with middle_panel:
                         voice_name=voice_name,
                         voice_rate=params.voice_rate,
                         voice_file=audio_file,
+                        voice_volume=params.voice_volume,
                     )
 
                 if sub_maker and os.path.exists(audio_file):
@@ -756,6 +765,32 @@ with middle_panel:
             config.azure["speech_region"] = azure_speech_region
             config.azure["speech_key"] = azure_speech_key
 
+        # 当选择硅基流动时，显示API key输入框和说明信息
+        if selected_tts_server == "siliconflow" or (
+            voice_name and voice.is_siliconflow_voice(voice_name)
+        ):
+            saved_siliconflow_api_key = config.siliconflow.get("api_key", "")
+
+            siliconflow_api_key = st.text_input(
+                tr("SiliconFlow API Key"),
+                value=saved_siliconflow_api_key,
+                type="password",
+                key="siliconflow_api_key_input",
+            )
+
+            # 显示硅基流动的说明信息
+            st.info(
+                tr("SiliconFlow TTS Settings")
+                + ":\n"
+                + "- "
+                + tr("Speed: Range [0.25, 4.0], default is 1.0")
+                + "\n"
+                + "- "
+                + tr("Volume: Uses Speech Volume setting, default 1.0 maps to gain 0")
+            )
+
+            config.siliconflow["api_key"] = siliconflow_api_key
+
         params.voice_volume = st.selectbox(
             tr("Speech Volume"),
             options=[0.6, 0.8, 1.0, 1.2, 1.5, 2.0, 3.0, 4.0, 5.0],

webui/i18n/de.json

@@ -93,6 +93,10 @@
     "TTS Provider": "Sprachsynthese-Anbieter auswählen",
     "TTS Servers": "TTS-Server",
     "No voices available for the selected TTS server. Please select another server.": "Keine Stimmen für den ausgewählten TTS-Server verfügbar. Bitte wählen Sie einen anderen Server.",
+    "SiliconFlow API Key": "SiliconFlow API-Schlüssel",
+    "SiliconFlow TTS Settings": "SiliconFlow TTS-Einstellungen",
+    "Speed: Range [0.25, 4.0], default is 1.0": "Geschwindigkeit: Bereich [0.25, 4.0], Standardwert ist 1.0",
+    "Volume: Uses Speech Volume setting, default 1.0 maps to gain 0": "Lautstärke: Verwendet die Sprachlautstärke-Einstellung, Standardwert 1.0 entspricht Verstärkung 0",
     "Hide Log": "Protokoll ausblenden",
     "Hide Basic Settings": "Basis-Einstellungen ausblenden\n\nWenn diese Option deaktiviert ist, wird die Basis-Einstellungen-Leiste nicht auf der Seite angezeigt.\n\nWenn Sie sie erneut anzeigen möchten, setzen Sie `hide_config = false` in `config.toml`",
     "LLM Settings": "**LLM-Einstellungen**",

webui/i18n/en.json

@@ -93,6 +93,10 @@
     "TTS Provider": "Select the voice synthesis provider",
     "TTS Servers": "TTS Servers",
     "No voices available for the selected TTS server. Please select another server.": "No voices available for the selected TTS server. Please select another server.",
+    "SiliconFlow API Key": "SiliconFlow API Key",
+    "SiliconFlow TTS Settings": "SiliconFlow TTS Settings",
+    "Speed: Range [0.25, 4.0], default is 1.0": "Speed: Range [0.25, 4.0], default is 1.0",
+    "Volume: Uses Speech Volume setting, default 1.0 maps to gain 0": "Volume: Uses Speech Volume setting, default 1.0 maps to gain 0",
     "Hide Log": "Hide Log",
     "Hide Basic Settings": "Hide Basic Settings\n\nHidden, the basic settings panel will not be displayed on the page.\n\nIf you need to display it again, please set `hide_config = false` in `config.toml`",
     "LLM Settings": "**LLM Settings**",

webui/i18n/pt.json

@@ -93,6 +93,10 @@
     "TTS Provider": "Selecione o provedor de síntese de voz",
     "TTS Servers": "Servidores TTS",
     "No voices available for the selected TTS server. Please select another server.": "Não há vozes disponíveis para o servidor TTS selecionado. Por favor, selecione outro servidor.",
+    "SiliconFlow API Key": "Chave API do SiliconFlow",
+    "SiliconFlow TTS Settings": "Configurações do SiliconFlow TTS",
+    "Speed: Range [0.25, 4.0], default is 1.0": "Velocidade: Intervalo [0.25, 4.0], o padrão é 1.0",
+    "Volume: Uses Speech Volume setting, default 1.0 maps to gain 0": "Volume: Usa a configuração de Volume de Fala, o padrão 1.0 corresponde ao ganho 0",
     "Hide Log": "Ocultar Log",
     "Hide Basic Settings": "Ocultar Configurações Básicas\n\nOculto, o painel de configurações básicas não será exibido na página.\n\nSe precisar exibi-lo novamente, defina `hide_config = false` em `config.toml`",
     "LLM Settings": "**Configurações do LLM**",

webui/i18n/vi.json

@@ -93,6 +93,10 @@
     "TTS Provider": "Chọn nhà cung cấp tổng hợp giọng nói",
     "TTS Servers": "Máy chủ TTS",
     "No voices available for the selected TTS server. Please select another server.": "Không có giọng nói nào cho máy chủ TTS đã chọn. Vui lòng chọn máy chủ khác.",
+    "SiliconFlow API Key": "Khóa API SiliconFlow",
+    "SiliconFlow TTS Settings": "Cài đặt SiliconFlow TTS",
+    "Speed: Range [0.25, 4.0], default is 1.0": "Tốc độ: Phạm vi [0.25, 4.0], mặc định là 1.0",
+    "Volume: Uses Speech Volume setting, default 1.0 maps to gain 0": "Âm lượng: Sử dụng cài đặt Âm lượng Giọng nói, mặc định 1.0 tương ứng với tăng ích 0",
     "Hide Log": "Ẩn Nhật Ký",
     "Hide Basic Settings": "Ẩn Cài Đặt Cơ Bản\n\nẨn, thanh cài đặt cơ bản sẽ không hiển thị trên trang web.\n\nNếu bạn muốn hiển thị lại, vui lòng đặt `hide_config = false` trong `config.toml`",
     "LLM Settings": "**Cài Đặt LLM**",

webui/i18n/zh.json

@@ -93,6 +93,10 @@
     "TTS Provider": "语音合成提供商",
     "TTS Servers": "TTS服务器",
     "No voices available for the selected TTS server. Please select another server.": "当前选择的TTS服务器没有可用的声音，请选择其他服务器。",
+    "SiliconFlow API Key": "硅基流动API密钥",
+    "SiliconFlow TTS Settings": "硅基流动TTS设置",
+    "Speed: Range [0.25, 4.0], default is 1.0": "语速范围 [0.25, 4.0]，默认值为1.0",
+    "Volume: Uses Speech Volume setting, default 1.0 maps to gain 0": "音量：使用朗读音量设置，默认值1.0对应增益0",
     "Hide Log": "隐藏日志",
     "Hide Basic Settings": "隐藏基础设置\n\n隐藏后，基础设置面板将不会显示在页面中。\n\n如需要再次显示，请在 `config.toml` 中设置 `hide_config = false`",
     "LLM Settings": "**大模型设置**",