Merge pull request Ikaros-521#996 from Ikaros-521/owner

新增：对接CosyVoice 0819的API

config.json

@@ -894,13 +894,20 @@
   "cosyvoice": {
     "type": "gradio_0707",
     "gradio_ip_port": "http://127.0.0.1:9886",
+    "api_ip_port": "http://127.0.0.1:9880",
     "gradio_0707": {
       "mode_checkbox_group": "预训练音色",
       "sft_dropdown": "中文女",
       "prompt_text": "",
       "prompt_wav_upload": "",
       "instruct_text": "",
       "seed": 0
+    },
+    "api_0819": {
+      "speaker": "中文女",
+      "new": 0,
+      "speed": 1.0,
+      "streaming": 0
     }
   },
   "choose_song": {

config.json.bak

@@ -894,13 +894,20 @@
   "cosyvoice": {
     "type": "gradio_0707",
     "gradio_ip_port": "http://127.0.0.1:9886",
+    "api_ip_port": "http://127.0.0.1:9880",
     "gradio_0707": {
       "mode_checkbox_group": "预训练音色",
       "sft_dropdown": "中文女",
       "prompt_text": "",
       "prompt_wav_upload": "",
       "instruct_text": "",
       "seed": 0
+    },
+    "api_0819": {
+      "speaker": "中文女",
+      "new": 0,
+      "speed": 1.0,
+      "streaming": 0
     }
   },
   "choose_song": {

tests/test_cosyvoice/api.py

@@ -0,0 +1,61 @@
+import json, logging, asyncio
+import aiohttp, requests, ssl
+from urllib.parse import urlencode
+import traceback
+from urllib.parse import urljoin
+
+async def download_audio(type: str, file_url: str, timeout: int=30, request_type: str="get", data=None, json_data=None, audio_suffix: str="wav"):
+    async with aiohttp.ClientSession() as session:
+        try:
+            if request_type == "get":
+                async with session.get(file_url, params=data, timeout=timeout) as response:
+                    if response.status == 200:
+                        content = await response.read()
+                        voice_tmp_path = '1.wav'
+                        with open(voice_tmp_path, 'wb') as file:
+                            file.write(content)
+                        return voice_tmp_path
+                    else:
+                        logging.error(f'{type} 下载音频失败: {response.status}')
+                        return None
+            else:
+                async with session.post(file_url, data=data, json=json_data, timeout=timeout) as response:
+                    if response.status == 200:
+                        content = await response.read()
+                        voice_tmp_path = '1.wav'
+                        with open(voice_tmp_path, 'wb') as file:
+                            file.write(content)
+                        return voice_tmp_path
+                    else:
+                        logging.error(f'{type} 下载音频失败: {response.status}')
+                        return None
+        except asyncio.TimeoutError:
+            logging.error("{type} 下载音频超时")
+            return None
+
+async def cosyvoice_api(text):
+    url = 'http://127.0.0.1:9880/'
+
+    params = {
+        "text": text,
+        "speaker": "中文女",
+        'new': 0,
+        'speed': 1.0,
+        'streaming': 0
+    }
+
+    logging.debug(f"params={params}")
+
+    try:
+        audio_path = await download_audio("cosyvoice", url, 30, request_type="post", json_data=params)
+        print(audio_path)
+        return audio_path
+    except Exception as e:
+        logging.error(traceback.format_exc())
+        logging.error(f'cosyvoice未知错误: {e}')
+
+    return None
+
+
+if __name__ == '__main__':
+    asyncio.run(cosyvoice_api("你好"))

utils/audio.py

@@ -1129,7 +1129,9 @@ async def tts_handle(self, message):
                 data = {
                     "type": message["data"]["type"],
                     "gradio_ip_port": message["data"]["gradio_ip_port"],
+                    "api_ip_port": message["data"]["api_ip_port"],
                     "gradio_0707": message["data"]["gradio_0707"],
+                    "api_0819": message["data"]["api_0819"],
                     "content": message["content"],
                 }
 
@@ -2084,7 +2086,9 @@ async def audio_synthesis_use_local_config(self, content, audio_synthesis_type="
             data = {
                 "type": self.config.get("cosyvoice", "type"),
                 "gradio_ip_port": self.config.get("cosyvoice", "gradio_ip_port"),
+                "api_ip_port": self.config.get("cosyvoice", "api_ip_port"),
                 "gradio_0707": self.config.get("cosyvoice", "gradio_0707"),
+                "api_0819": self.config.get("cosyvoice", "api_0819"),
                 "content": content
             }
             # 调用接口合成语音

utils/audio_handle/my_tts.py

@@ -1295,6 +1295,26 @@ async def cosyvoice_api(self, data):
                     new_file_path = self.common.move_file(voice_tmp_path, os.path.join(self.audio_out_path, 'cosyvoice_' + self.common.get_bj_time(4)), 'cosyvoice_' + self.common.get_bj_time(4))
 
                 return new_file_path
+            elif data["type"] == "api_0819":
+                url = data["api_ip_port"]
+
+                params = {
+                    "text": data["content"],
+                    "speaker": data["api_0819"]["speaker"],
+                    'new': int(data["api_0819"]["new"]),
+                    'speed': float(data["api_0819"]["speed"]),
+                    'streaming': int(data["api_0819"]["streaming"])
+                }
+
+                logger.debug(f"params={params}")
+
+                try:
+                    return await self.download_audio("cosyvoice", url, self.timeout, request_type="post", json_data=params)
+                except Exception as e:
+                    logger.error(traceback.format_exc())
+                    logger.error(f'cosyvoice未知错误，请检查您的CosyVoice API是否启动/配置是否正确，报错内容: {e}')
+
+                return None
         except Exception as e:
             logger.error(traceback.format_exc())
             logger.error(f'CosyVoice未知错误，请检查您的CosyVoice WebUI是否启动/配置是否正确，报错内容: {e}')

webui.py

@@ -2518,13 +2518,18 @@ def common_textarea_handle(content):
                 if config.get("webui", "show_card", "tts", "cosyvoice"):
                     config_data["cosyvoice"]["type"] = select_cosyvoice_type.value
                     config_data["cosyvoice"]["gradio_ip_port"] = input_cosyvoice_gradio_ip_port.value
+                    config_data["cosyvoice"]["api_ip_port"] = input_cosyvoice_api_ip_port.value
                     config_data["cosyvoice"]["gradio_0707"]["mode_checkbox_group"] = select_cosyvoice_gradio_0707_mode_checkbox_group.value
                     config_data["cosyvoice"]["gradio_0707"]["sft_dropdown"] = select_cosyvoice_gradio_0707_sft_dropdown.value
                     config_data["cosyvoice"]["gradio_0707"]["prompt_text"] = input_cosyvoice_gradio_0707_prompt_text.value
                     config_data["cosyvoice"]["gradio_0707"]["prompt_wav_upload"] = input_cosyvoice_gradio_0707_prompt_wav_upload.value
                     config_data["cosyvoice"]["gradio_0707"]["instruct_text"] = input_cosyvoice_gradio_0707_instruct_text.value
                     config_data["cosyvoice"]["gradio_0707"]["seed"] = int(input_cosyvoice_gradio_0707_seed.value)
 
+                    config_data["cosyvoice"]["api_0819"]["speaker"] =  input_cosyvoice_api_0819_speaker.value
+                    config_data["cosyvoice"]["api_0819"]["new"] = int(input_cosyvoice_api_0819_new.value)
+                    config_data["cosyvoice"]["api_0819"]["speed"] = round(float(input_cosyvoice_api_0819_speed.value), 2)
+
             """
             SVC
             """
@@ -5582,7 +5587,7 @@ async def fish_speech_load_model(data):
                     with ui.row():
                         select_cosyvoice_type = ui.select(
                             label='类型', 
-                            options={"gradio_0707": "gradio_0707"}, 
+                            options={"api_0819": "api_0819", "gradio_0707": "gradio_0707"}, 
                             value=config.get("cosyvoice", "type")
                         ).style("width:150px").tooltip("对接的API类型")
                         input_cosyvoice_gradio_ip_port = ui.input(
@@ -5593,9 +5598,18 @@ async def fish_speech_load_model(data):
                                 '请输入正确格式的URL': lambda value: common.is_url_check(value),
                             }
                         ).style("width:200px;").tooltip("对接webui的gradio接口，填webui的地址")
+                        input_cosyvoice_api_ip_port = ui.input(
+                            label='HTTP API地址', 
+                            value=config.get("cosyvoice", "api_ip_port"), 
+                            placeholder='API程序启动后，API请求地址',
+                            validation={
+                                '请输入正确格式的URL': lambda value: common.is_url_check(value),
+                            }
+                        ).style("width:200px;").tooltip("对接api接口，填api端点地址")
 
                     with ui.row():
                         with ui.card().style(card_css):
+                            ui.label("gradio_0707")
                             with ui.row():
                                 select_cosyvoice_gradio_0707_mode_checkbox_group = ui.select(
                                     label='推理模式', 
@@ -5611,7 +5625,14 @@ async def fish_speech_load_model(data):
                                 input_cosyvoice_gradio_0707_prompt_wav_upload = ui.input(label='prompt音频路径', value=config.get("cosyvoice", "gradio_0707", "prompt_wav_upload"), placeholder='例如：E:\\1.wav').style("width:200px;").tooltip("不用就留空，例如：E:\\1.wav")
                                 input_cosyvoice_gradio_0707_instruct_text = ui.input(label='instruct文本', value=config.get("cosyvoice", "gradio_0707", "instruct_text"), placeholder='').style("width:200px;").tooltip("不用就留空")
                                 input_cosyvoice_gradio_0707_seed = ui.input(label='随机推理种子', value=config.get("cosyvoice", "gradio_0707", "seed"), placeholder='默认：0').style("width:100px;").tooltip("随机推理种子")
-
+                    with ui.row():
+                        with ui.card().style(card_css):
+                            ui.label("api_0819")
+                            with ui.row():
+                                input_cosyvoice_api_0819_speaker = ui.input(label='说话人', value=config.get("cosyvoice", "api_0819", "speaker"), placeholder='').style("width:200px;").tooltip("自行查看")
+                                input_cosyvoice_api_0819_new = ui.input(label='new', value=config.get("cosyvoice", "api_0819", "new"), placeholder='0').style("width:200px;").tooltip("自行查看")
+                                input_cosyvoice_api_0819_speed = ui.input(label='语速', value=config.get("cosyvoice", "api_0819", "speed"), placeholder='1').style("width:200px;").tooltip("语速")
+
         with ui.tab_panel(svc_page).style(tab_panel_css):
             if config.get("webui", "show_card", "svc", "ddsp_svc"):
                 with ui.card().style(card_css):