Merge pull request #78 from sudoskys/tts_azure_vits

TTS 异步支持，添加网络请求库，AZURE 添加
LlmKira · Dec 20, 2022 · 6d62d99 · 6d62d99
2 parents af0272d + 89cbca4
commit 6d62d99
Show file tree

Hide file tree

Showing 7 changed files with 354 additions and 149 deletions.
diff --git a/App/Event.py b/App/Event.py
@@ -69,38 +69,64 @@ def save_csonfig():
         json.dump(_csonfig, f, indent=4, ensure_ascii=False)
 
 
-def TTS_Support_Check(text, user_id):
+async def TTS_Support_Check(text, user_id):
     global _tts_conf
     """
     处理消息文本并构造请求返回字节流或者空。隶属 Event 文件
     :return:
     """
+    from openai_async.utils.Talk import Talk
     if not _tts_conf["status"]:
         return
+    lang_type = Talk.get_language(text)
+    if not lang_type in ["chinese"]:
+        return
     if _tts_conf["type"] == "vits":
-        from openai_async.utils.Talk import Talk
-        if Talk.get_language(text) != "chinese":
-            return
-        if len(text) > _tts_conf["vits"]["limit"]:
+        _vits_config = _tts_conf["vits"]
+        if len(text) > _vits_config["limit"]:
             return
+        # 简单处理文本
         _new_text = f"[ZH]{text}[ZH]"
-        result = TTS_Clint.request_vits_server(url=_tts_conf["vits"]["api"],
-                                               params=TTS_REQ(task_id=user_id,
-                                                              text=_new_text,
-                                                              model_name=_tts_conf["vits"]["model_name"],
-                                                              speaker_id=_tts_conf["vits"]["speaker_id"]))
+        # 接受数据
+        result, e = await TTS_Clint.request_vits_server(url=_vits_config["api"],
+                                                        params=TTS_REQ(task_id=user_id,
+                                                                       text=_new_text,
+                                                                       model_name=_vits_config["model_name"],
+                                                                       speaker_id=_vits_config["speaker_id"]))
         if not result:
-            logger.error(f"RUN:{user_id} --vits: {text}:{len(text)} TTS failed")
+            logger.error(f"TTS:{user_id} --type:vits --content: {text}:{len(text)} --{e}")
             return
-        try:
-            data = TTS_Clint.decode_wav(result["audio"])
-        except:
-            logger.error(f"RUN:{user_id} --vits: {text}:{len(text)} decode tts data failed")
+        logger.info(f"TTS:{user_id} --type:vits --content: {text}:{len(text)}")
+        # 返回字节流
+        return result
+    # USE AZURE
+    elif _tts_conf["type"] == "azure":
+        _azure_config = _tts_conf["azure"]
+        _new_text = text
+        _speaker = _azure_config["speaker"].get(lang_type)
+
+        if len(text) > _azure_config["limit"]:
             return
-        else:
-            logger.info(f"RUN:{user_id} --vits: {text}:{len(text)}")
-            return data
-    return
+
+        if not _speaker:
+            logger.info(f"TTS:{user_id} --type:azure --content: {text}:{len(text)} --this type lang not supported")
+            return
+
+        result, e = await TTS_Clint.request_azure_server(key=_azure_config["key"],
+                                                         location=_azure_config["location"],
+                                                         text=_new_text,
+                                                         speaker=_speaker
+                                                         )
+        if not result:
+            logger.error(f"TTS:{user_id} --type:azure --content: {text}:{len(text)} --{e}")
+            return
+
+        logger.info(f"TTS:{user_id} --type:azure --content: {text}:{len(text)}")
+        # 返回字节流
+        return result
+    else:
+        # 啥也没有
+        return
 
 
 async def Forget(bot, message, config):
@@ -354,7 +380,7 @@ async def Text(bot, message, config, reset: bool = False):
                                          start_name="Girl:",
                                          web_enhance_server=config.Enhance_Server
                                          )
-        voice = TTS_Support_Check(text=_req, user_id=message.from_user.id)
+        voice = await TTS_Support_Check(text=_req, user_id=message.from_user.id)
         if voice:
             if not message.chat.is_forum:
                 msg = await bot.send_audio(chat_id=message.chat.id, audio=voice, title="Voice")
@@ -413,7 +439,7 @@ async def private_Chat(bot, message, config):
                                              method=types,
                                              web_enhance_server=config.Enhance_Server
                                              )
-            voice = TTS_Support_Check(text=_req, user_id=message.from_user.id)
+            voice = await TTS_Support_Check(text=_req, user_id=message.from_user.id)
             if voice:
                 if not message.chat.is_forum:
                     msg = await bot.send_audio(chat_id=message.chat.id, audio=voice, title="Voice")

diff --git a/README.EN.md b/README.EN.md
@@ -186,14 +186,47 @@ that are not in the preset will not be completed.
     "type": "vits",
     "vits": {
       "api": "http://127.0.0.1:9557/tts/generate",
-      "limit": 20,
+      "limit": 70,
       "model_name": "some.pth",
-      "speaker_id": 1
+      "speaker_id": 0
+    },
+    "azure": {
+      "key": [
+        "123"
+      ],
+      "limit": 70,
+      "speaker": {
+        "chinese": "zh-CN-XiaoxiaoNeural"
+      },
+      "location": "japanwest"
     }
   }
 }
 ```
 
+**Redis**
+
+- slightly
+
+**TTS**
+
+- status switch
+- type Type
+
+*VITS*
+
+- vits:limit text within length will be converted
+- vits:model_name The name of the model, some.pth, in the model folder
+- vits:speaker_id The ID of the speaker, see the model config
+
+*Azure*
+
+- azure:limit The text within the length will be converted
+- azure:
+  speaker[list-of-all-sound-engines](https://learn.microsoft.com/en-us/azure/cognitive-services/speech-service/language-support?tabs=stt-tts)
+- auzre:location Server resource address
+- auzre:key api key
+
 #### VITS Voice Support Description(ONLY CN NOW)
 
 This technology provides an emulated voice interaction technique.

diff --git a/README.md b/README.md
@@ -185,37 +185,62 @@ add_api_key - 增加 Api key
     "type": "vits",
     "vits": {
       "api": "http://127.0.0.1:9557/tts/generate",
-      "limit": 20,
+      "limit": 70,
       "model_name": "some.pth",
-      "speaker_id": 1
+      "speaker_id": 0
+    },
+    "azure": {
+      "key": [
+        "123"
+      ],
+      "limit": 70,
+      "speaker": {
+        "chinese": "zh-CN-XiaoxiaoNeural"
+      },
+      "location": "japanwest"
     }
   }
 }
 ```
 
-redis:
+**Redis**
 
-- ?
+- 略
 
-tts:
+**TTS**
 
 - status 开关
 - type 类型
+
+#### Azure 支持说明
+
+https://azure.microsoft.com/en-us/pricing/details/cognitive-services/speech-services/
+
+*Azure*
+
+- azure:limit 长度内的文本会被转换
+- azure:speaker
+  说话人, [所有声音引擎列表](https://learn.microsoft.com/en-us/azure/cognitive-services/speech-service/language-support?tabs=stt-tts)
+- auzre:location 服务器资源地址
+- auzre:key api key
+
+#### VITS 语音支持说明(ONLY CN NOW)
+
+Api 后端请使用我打包改造的 MoeGoe https://github.com/sudoskys/MoeGoe 本机运行
+
+*VITS*
+
 - vits:limit 长度内的文本会被转换
 - vits:model_name 模型名字，some.pth,在 model 文件夹下的
 - vits:speaker_id 说话人的ID,具体看模型config
 
-#### VITS 语音支持说明(ONLY CN NOW)
-
 这项技术提供了一种仿真的语音交互技术。
 
-Api 后端为我打包改造的 MoeGoe https://github.com/sudoskys/MoeGoe
-
 安装依赖，运行 `server.py` 文件可以默认使用。
 
-模型下面请查询 MoeGoe 项目的 Readme,并注意模型相应的协议。
-
-如果不说话，可能是文本的长度超过了设定的 limit。长度为直接取 `len()`
+模型下载请查询 MoeGoe 项目的 Readme,并注意模型相应的协议。
+如果不生效，可能是文本的长度超过了设定的 limit 。
+长度为直接取 `len()`
 
 ## 运行
 

diff --git a/openai_async/api/network.py b/openai_async/api/network.py
@@ -98,8 +98,8 @@ def get_session(proxy: str = ""):
     loop = asyncio.get_event_loop()
     session = __session_pool.get(loop, None)
     if session is None:
-        if proxy != "":
-            proxies = {"all": proxy}
+        if proxy:
+            proxies = {"all://": proxy}
             session = httpx.AsyncClient(timeout=300, proxies=proxies)
         else:
             session = httpx.AsyncClient(timeout=300)