bert_vits2 新增刘悦-中文特化API 的对接

qymh233 · Jun 14, 2024 · 71e7bfc · 71e7bfc
1 parent 46b7e8e
commit 71e7bfc
Show file tree

Hide file tree

Showing 5 changed files with 215 additions and 107 deletions.
diff --git a/config.json b/config.json
@@ -615,7 +615,23 @@
     "auto_split": false,
     "emotion": "",
     "style_text": "",
-    "style_weight": 0.7
+    "style_weight": 0.7,
+    "刘悦-中文特化API": {
+      "api_ip_port": "http://127.0.0.1:9885",
+      "speaker": "Yennefer",
+      "language": "ZH",
+      "length_scale": "0.9-1.1",
+      "noise_scale": "0.2-0.3",
+      "noise_scale_w": "0.85-0.95",
+      "sdp_radio": "0.2-0.3",
+      "cut_by_sent": true,
+      "interval_between_para": "1",
+      "interval_between_sent": "0.2",
+      "emotion": "Happy",
+      "style_text": "",
+      "style_weight": "0.7-0.8",
+      "stream": false
+    }
   },
   "vits_fast": {
     "config_path": "D:\\GitHub_pro\\VITS-fast-fine-tuning\\inference\\finetune_speaker.json",
@@ -1782,10 +1798,10 @@
       "visual_body": {
         "live2d": true,
         "xuniren": true,
-        "metahuman": true,
-        "musetalk": true,
         "unity": true,
         "EasyAIVtuber": true,
+        "metahuman": true,
+        "musetalk": true,
         "digital_human_video_player": true
       }
     }

diff --git a/config.json.bak b/config.json.bak
@@ -615,7 +615,23 @@
     "auto_split": false,
     "emotion": "",
     "style_text": "",
-    "style_weight": 0.7
+    "style_weight": 0.7,
+    "刘悦-中文特化API": {
+      "api_ip_port": "http://127.0.0.1:9885",
+      "speaker": "Yennefer",
+      "language": "ZH",
+      "length_scale": "0.9-1.1",
+      "noise_scale": "0.2-0.3",
+      "noise_scale_w": "0.85-0.95",
+      "sdp_radio": "0.2-0.3",
+      "cut_by_sent": true,
+      "interval_between_para": "1",
+      "interval_between_sent": "0.2",
+      "emotion": "Happy",
+      "style_text": "",
+      "style_weight": "0.7-0.8",
+      "stream": false
+    }
   },
   "vits_fast": {
     "config_path": "D:\\GitHub_pro\\VITS-fast-fine-tuning\\inference\\finetune_speaker.json",

diff --git a/utils/audio.py b/utils/audio.py
@@ -844,46 +844,46 @@ async def tts_handle(self, message):
                 voice_tmp_path = await self.my_tts.vits_api(data)
 
             elif message["tts_type"] == "bert_vits2":
-                if message["data"]["type"] == "hiyori":
-                    if message["data"]["language"] == "auto":
-                        # 自动检测语言
-                        language = self.common.lang_check(message["content"])
+                if message["data"]["language"] == "auto":
+                    # 自动检测语言
+                    language = self.common.lang_check(message["content"])
 
-                        logging.debug(f'language={language}')
+                    logging.debug(f'language={language}')
 
-                        # 自定义语言名称（需要匹配请求解析）
-                        language_name_dict = {"en": "EN", "zh": "ZH", "ja": "JP"}  
+                    # 自定义语言名称（需要匹配请求解析）
+                    language_name_dict = {"en": "EN", "zh": "ZH", "ja": "JP"}  
 
-                        if language in language_name_dict:
-                            language = language_name_dict[language]
-                        else:
-                            language = "ZH"  # 无法识别出语言代码时的默认值
+                    if language in language_name_dict:
+                        language = language_name_dict[language]
                     else:
-                        language = message["data"]["language"]
+                        language = "ZH"  # 无法识别出语言代码时的默认值
+                else:
+                    language = message["data"]["language"]
 
-                    data = {
-                        "api_ip_port": message["data"]["api_ip_port"],
-                        "type": message["data"]["type"],
-                        "model_id": message["data"]["model_id"],
-                        "speaker_name": message["data"]["speaker_name"],
-                        "speaker_id": message["data"]["speaker_id"],
-                        "language": language,
-                        "length": message["data"]["length"],
-                        "noise": message["data"]["noise"],
-                        "noisew": message["data"]["noisew"],
-                        "sdp_radio": message["data"]["sdp_radio"],
-                        "auto_translate": message["data"]["auto_translate"],
-                        "auto_split": message["data"]["auto_split"],
-                        "emotion": message["data"]["emotion"],
-                        "style_text": message["data"]["style_text"],
-                        "style_weight": message["data"]["style_weight"],
-                        "content": message["content"]
-                    }
+                data = {
+                    "api_ip_port": message["data"]["api_ip_port"],
+                    "type": message["data"]["type"],
+                    "model_id": message["data"]["model_id"],
+                    "speaker_name": message["data"]["speaker_name"],
+                    "speaker_id": message["data"]["speaker_id"],
+                    "language": language,
+                    "length": message["data"]["length"],
+                    "noise": message["data"]["noise"],
+                    "noisew": message["data"]["noisew"],
+                    "sdp_radio": message["data"]["sdp_radio"],
+                    "auto_translate": message["data"]["auto_translate"],
+                    "auto_split": message["data"]["auto_split"],
+                    "emotion": message["data"]["emotion"],
+                    "style_text": message["data"]["style_text"],
+                    "style_weight": message["data"]["style_weight"],
+                    "刘悦-中文特化API": message["data"]["刘悦-中文特化API"],
+                    "content": message["content"]
+                }
 
-
 
                 # 调用接口合成语音
                 voice_tmp_path = await self.my_tts.bert_vits2_api(data)
+
             elif message["tts_type"] == "vits_fast":
                 if message["data"]["language"] == "自动识别":
                     # 自动检测语言
@@ -1768,46 +1768,47 @@ async def audio_synthesis_use_local_config(self, content, audio_synthesis_type="
 
 
         elif audio_synthesis_type == "bert_vits2":
-            if self.config.get("bert_vits2", "type") == "hiyori":
-                if self.config.get("bert_vits2", "language") == "auto":
-                    # 自动检测语言
-                    language = self.common.lang_check(content)
+
+            if self.config.get("bert_vits2", "language") == "auto":
+                # 自动检测语言
+                language = self.common.lang_check(content)
 
-                    logging.debug(f'language={language}')
+                logging.debug(f'language={language}')
 
-                    # 自定义语言名称（需要匹配请求解析）
-                    language_name_dict = {"en": "EN", "zh": "ZH", "ja": "JP"}  
+                # 自定义语言名称（需要匹配请求解析）
+                language_name_dict = {"en": "EN", "zh": "ZH", "ja": "JP"}  
 
-                    if language in language_name_dict:
-                        language = language_name_dict[language]
-                    else:
-                        language = "ZH"  # 无法识别出语言代码时的默认值
+                if language in language_name_dict:
+                    language = language_name_dict[language]
                 else:
-                    language = self.config.get("bert_vits2", "language")
-
-                data = {
-                    "api_ip_port": self.config.get("bert_vits2", "api_ip_port"),
-                    "type": self.config.get("bert_vits2", "type"),
-                    "model_id": self.config.get("bert_vits2", "model_id"),
-                    "speaker_name": self.config.get("bert_vits2", "speaker_name"),
-                    "speaker_id": self.config.get("bert_vits2", "speaker_id"),
-                    "language": language,
-                    "length": self.config.get("bert_vits2", "length"),
-                    "noise": self.config.get("bert_vits2", "noise"),
-                    "noisew": self.config.get("bert_vits2", "noisew"),
-                    "sdp_radio": self.config.get("bert_vits2", "sdp_radio"),
-                    "auto_translate": self.config.get("bert_vits2", "auto_translate"),
-                    "auto_split": self.config.get("bert_vits2", "auto_split"),
-                    "emotion": self.config.get("bert_vits2", "emotion"),
-                    "style_text": self.config.get("bert_vits2", "style_text"),
-                    "style_weight": self.config.get("bert_vits2", "style_weight"),
-                    "content": content
-                }
+                    language = "ZH"  # 无法识别出语言代码时的默认值
+            else:
+                language = self.config.get("bert_vits2", "language")
+
+            data = {
+                "api_ip_port": self.config.get("bert_vits2", "api_ip_port"),
+                "type": self.config.get("bert_vits2", "type"),
+                "model_id": self.config.get("bert_vits2", "model_id"),
+                "speaker_name": self.config.get("bert_vits2", "speaker_name"),
+                "speaker_id": self.config.get("bert_vits2", "speaker_id"),
+                "language": language,
+                "length": self.config.get("bert_vits2", "length"),
+                "noise": self.config.get("bert_vits2", "noise"),
+                "noisew": self.config.get("bert_vits2", "noisew"),
+                "sdp_radio": self.config.get("bert_vits2", "sdp_radio"),
+                "auto_translate": self.config.get("bert_vits2", "auto_translate"),
+                "auto_split": self.config.get("bert_vits2", "auto_split"),
+                "emotion": self.config.get("bert_vits2", "emotion"),
+                "style_text": self.config.get("bert_vits2", "style_text"),
+                "style_weight": self.config.get("bert_vits2", "style_weight"),
+                "刘悦-中文特化API": self.config.get("bert_vits2", "刘悦-中文特化API"),
+                "content": content
+            }
 
-                logging.info(f"data={data}")
+            logging.info(f"data={data}")
 
-                # 调用接口合成语音
-                voice_tmp_path = await self.my_tts.bert_vits2_api(data)
+            # 调用接口合成语音
+            voice_tmp_path = await self.my_tts.bert_vits2_api(data)
         elif audio_synthesis_type == "vits_fast":
             if vits_fast["language"] == "自动识别":
                 # 自动检测语言

diff --git a/utils/audio_handle/my_tts.py b/utils/audio_handle/my_tts.py
@@ -256,14 +256,42 @@ async def bert_vits2_api(self, data):
                     "style_weight": self.get_random_float(data["style_weight"])
                 }
 
-            logging.debug(f"data_json={data_json}")
-            # logging.info(f"data={data}")
+                logging.debug(f"data_json={data_json}")
+                # logging.info(f"data={data}")
 
-            logging.debug(f"API_URL={API_URL}")
+                logging.debug(f"API_URL={API_URL}")
 
-            url = f"{API_URL}?{urlencode(data_json)}"
+                url = f"{API_URL}?{urlencode(data_json)}"
+
+                return await self.download_audio("bert_vits2", url, self.timeout)
+            elif data["type"] == "刘悦-中文特化API":
+                type = data["type"]
+                # API地址 "http://127.0.0.1:5000/run/predict/"
+                API_URL = urljoin(data[type]["api_ip_port"], '/tts_to_audio/')
+
+                data_json = {
+                    "text": data["content"],
+                    "speaker": data[type]["speaker"],
+                    "language": data["language"],
+                    "length_scale": self.get_random_float(data[type]["length_scale"]),
+                    "noise_scale": self.get_random_float(data[type]["noise_scale"]),
+                    "noise_scale_w": self.get_random_float(data[type]["noise_scale_w"]),
+                    "sdp_radio": self.get_random_float(data[type]["sdp_radio"]),
+                    "cut_by_sent": data[type]["cut_by_sent"],
+                    "interval_between_para": self.get_random_float(data[type]["interval_between_para"]),
+                    "interval_between_sent": self.get_random_float(data[type]["interval_between_sent"]),
+                    "emotion": data[type]["emotion"],
+                    "style_text": data[type]["style_text"],
+                    "style_weight": self.get_random_float(data[type]["style_weight"]),
+                    "stream": data[type]["stream"]
+                }
+
+                logging.debug(f"data_json={data_json}")
+                # logging.info(f"data={data}")
+
+                logging.debug(f"API_URL={API_URL}")
 
-            return await self.download_audio("bert_vits2", url, self.timeout)
+                return await self.download_audio("bert_vits2", API_URL, self.timeout, "post", json_data=data_json)
         except aiohttp.ClientError as e:
             logging.error(traceback.format_exc())
             logging.error(f'bert_vits2请求失败，请检查您的bert_vits2 api是否启动/配置是否正确，报错内容: {e}')