添加chunk处理

Human.py

@@ -1,14 +1,50 @@
 #encoding = utf8
+import logging
+import queue
 from queue import Queue
 
+import numpy as np
+
+from tts.Chunk2Mal import Chunk2Mal
+
+logger = logging.getLogger(__name__)
 
 class Human:
     def __init__(self):
         self._tts = None
-        self._audio_chunk_queue = Queue()
+        self._fps = 25  # 20 ms per frame
+        self._sample_rate = 16000
+        self._chunk = self._sample_rate // self._fps  # 320 samples per chunk (20ms * 16000 / 1000)
+        self._chunk_2_mal = Chunk2Mal()
+
+    def on_destroy(self):
+        if self._tts is not None:
+            self._tts.stop()
+        logger.info('human destroy')
 
     def set_tts(self, tts):
+        if self._tts == tts:
+            return
+
         self._tts = tts
+        self._tts.start()
+
+    def read(self, txt):
+        if self._tts is None:
+            logger.warning('tts is none')
+            return
+
+        self._tts.push_txt(txt)
 
     def push_audio_chunk(self, chunk):
-        pass
+        self._chunk_2_mal.push_chunk(chunk)
+        self._audio_chunk_queue.put(chunk)
+
+    def pull_audio_chunk(self):
+        try:
+            chunk = self._audio_chunk_queue.get(block=True, timeout=1.0)
+            type = 1
+        except queue.Empty:
+            chunk = np.zeros(self._chunk, dtype=np.float32)
+            type = 0
+        return chunk, type

requirements.txt

@@ -2,7 +2,7 @@ librosa~=0.10.2.post1
 numpy~=1.26.3
 opencv-contrib-python
 opencv-python~=4.10.0.84
-torch~=2.4.0+cu118
+torch
 torchvision
 tqdm~=4.66.5
 numba

tts/Chunk2Mal.py

@@ -0,0 +1,11 @@
+#encoding = utf8
+from queue import Queue
+
+
+class Chunk2Mal:
+    def __init__(self):
+        self._audio_chunk_queue = Queue()
+
+    def push_chunk(self, chunk):
+        self._audio_chunk_queue.put(chunk)
+

tts/TTSBase.py

@@ -1,4 +1,5 @@
 #encoding = utf8
+import logging
 import queue
 from io import BytesIO
 from queue import Queue
@@ -17,12 +18,14 @@ class TTSBase:
         self._chunk = self._sample_rate // self._fps
 
     def _on_run(self):
+        logging.info('tts run')
         while not self._exit_event.is_set():
             try:
                 txt = self._queue.get(block=True, timeout=1)
             except queue.Empty:
                 continue
             self._request(txt)
+        logging.info('tts exit')
 
     def _request(self, txt):
         pass
@@ -33,6 +36,7 @@ class TTSBase:
         self._exit_event = Event()
         self._thread = Thread(target=self._on_run)
         self._thread.start()
+        logging.info('tts start')
 
     def stop(self):
         if self._exit_event is None:
@@ -40,6 +44,7 @@ class TTSBase:
 
         self._exit_event.set()
         self._thread.join()
+        logging.info('tts stop')
 
     def clear(self):
         self._queue.queue.clear()

ui.py

@@ -1,6 +1,7 @@
 #encoding = utf8
 import json
 import logging
+from logging import handlers
 import tkinter
 import tkinter.messagebox
 import customtkinter
@@ -52,6 +53,10 @@ class App(customtkinter.CTk):
         tts = EdgeTTS(self._human)
         self._human.set_tts(tts)
 
+    def on_destroy(self):
+        logger.info('------------App destroy------------')
+        self._human.on_destroy()
+
     def _init_image_canvas(self):
         self._canvas = customtkinter.CTkCanvas(self.image_frame)
         self._canvas.pack(fill=customtkinter.BOTH, expand=customtkinter.YES)
@@ -60,20 +65,21 @@ class App(customtkinter.CTk):
         content = self.entry.get()
         print('content:', content)
         self.entry.delete(0, customtkinter.END)
-        payload = {
-          'text': content,
-          'voice': 'zh-CN-XiaoyiNeural'
-        }
-        resp = requests.get(self._tts_url + '/tts', params=urlencode(payload))
-        if resp.status_code != 200:
-            print('tts error', resp.status_code)
-            return
-
-        print(resp.content)
-
-        resJson = json.loads(resp.text)
-        url = resJson.get('url')
-        self.download_tts(url)
+        self._human.read(content)
+        # payload = {
+        #   'text': content,
+        #   'voice': 'zh-CN-XiaoyiNeural'
+        # }
+        # resp = requests.get(self._tts_url + '/tts', params=urlencode(payload))
+        # if resp.status_code != 200:
+        #     print('tts error', resp.status_code)
+        #     return
+        #
+        # print(resp.content)
+        #
+        # resJson = json.loads(resp.text)
+        # url = resJson.get('url')
+        # self.download_tts(url)
 
     def download_tts(self, url):
         file_name = url[3:]
@@ -91,8 +97,43 @@ class App(customtkinter.CTk):
         # open('./audio/', 'wb') with
 
 
-if __name__ == "__main__":
-    logging.basicConfig(filename='./logs/info.log', level=logging.INFO)
+def config_logging(file_name: str, console_level: int=logging.INFO, file_level: int=logging.DEBUG):
+    file_handler = logging.FileHandler(file_name, mode='a', encoding="utf8")
+    file_handler.setFormatter(logging.Formatter(
+        '%(asctime)s [%(levelname)s] %(module)s.%(lineno)d %(name)s:\t%(message)s'
+        ))
+    file_handler.setLevel(file_level)
 
+    console_handler = logging.StreamHandler()
+    console_handler.setFormatter(logging.Formatter(
+        '[%(asctime)s %(levelname)s] %(message)s',
+        datefmt="%Y/%m/%d %H:%M:%S"
+        ))
+    console_handler.setLevel(console_level)
+
+    logging.basicConfig(
+        level=min(console_level, file_level),
+        handlers=[file_handler, console_handler],
+        )
+
+
+if __name__ == "__main__":
+    # logging.basicConfig(filename='./logs/info.log', level=logging.INFO)
+    config_logging('./logs/info.log', logging.INFO, logging.INFO)
+    # logger = logging.getLogger('manager')
+    # # 输出到控制台, 级别为DEBUG
+    # console = logging.StreamHandler()
+    # console.setLevel(logging.DEBUG)
+    # logger.addHandler(console)
+    #
+    # # 输出到文件, 级别为INFO, 文件按大小切分
+    # filelog = logging.handlers.RotatingFileHandler(filename='./logs/info.log', level=logging.INFO,
+    #                                                maxBytes=1024 * 1024, backupCount=5)
+    # filelog.setFormatter(logging.Formatter('%(asctime)s - %(levelname)s - %(message)s'))
+    # logger.setLevel(logging.INFO)
+    # logger.addHandler(filelog)
+    logger.info('------------start------------')
     app = App()
     app.mainloop()
+    app.on_destroy()
+    # logger.info('------------exit------------')