modify handle

audio_render/__init__.py

@@ -1,3 +0,0 @@
-#encoding = utf8
-
-from .audio_render import AudioRender

audio_render/audio_render.py

@@ -1,39 +0,0 @@
-#encoding = utf8
-
-from ctypes import *
-import os
-
-import numpy as np
-
-current = os.path.dirname(__file__)
-dynamic_path = os.path.join(current, 'AudioRender.dll')
-
-
-def audio_render_log_callback(level, log, size):
-    print(f'level={level}, log={log}, len={size}')
-
-
-class AudioRender:
-    def __init__(self):
-        self.__audio_render_obj = WinDLL(dynamic_path)
-        print(self.__audio_render_obj)
-        if self.__audio_render_obj is not None:
-            CALLBACK_TYPE = CFUNCTYPE(None, c_int, c_ubyte, c_uint)
-            c_callback = CALLBACK_TYPE(audio_render_log_callback)
-            self.__init = self.__audio_render_obj.Initialize(c_callback)
-            print('AudioRender init', self.__init)
-
-    def __del__(self):
-        print('AudioRender __del__')
-        if self.__audio_render_obj is None:
-            return
-        if self.__init:
-            self.__audio_render_obj.Uninitialize()
-
-    def write(self, data, size):
-        if not self.__init:
-            return False
-
-        self.__audio_render_obj.argtypes = (POINTER(c_uint8), c_uint)
-        byte_data = np.frombuffer(data, dtype=np.uint8)
-        return self.__audio_render_obj.Write(byte_data.ctypes.data_as(POINTER(c_uint8)), size)

human/audio_inference_handler.py

@@ -103,10 +103,10 @@ class AudioInferenceHandler(AudioHandler):
                     for i in range(batch_size):
                         if not self._is_running:
                             break
-                        self.on_next_handle((None, mirror_index(length, index),
+                        # self.on_next_handle((None, mirror_index(length, index),
-                        # self.on_next_handle((None, human_status.get_index(),
+                        self.on_next_handle((None, human_status.get_index(),
                                              audio_frames[i * 2:i * 2 + 2]), 0)
-                        index = index + 1
+                        # index = index + 1
                 else:
                     human_status.start_talking()
                     logger.info(f'infer======= {current_text}')
@@ -116,9 +116,9 @@ class AudioInferenceHandler(AudioHandler):
                     index_list = []
                     # for i in range(batch_size):
                     for i in range(len(mel_batch)):
-                        idx = mirror_index(length, index + i)
+                        # idx = mirror_index(length, index + i)
-                        # idx = human_status.get_index()
+                        idx = human_status.get_index()
-                        # index_list.append(idx)
+                        index_list.append(idx)
                         face = face_list_cycle[idx]
                         img_batch.append(face)
 
@@ -152,10 +152,10 @@ class AudioInferenceHandler(AudioHandler):
                         if not self._is_running:
                             break
                         self.on_next_handle(
-                            (res_frame, mirror_index(length, index), audio_frames[i * 2:i * 2 + 2]),
+                            # (res_frame, mirror_index(length, index), audio_frames[i * 2:i * 2 + 2]),
-                            # (res_frame, index_list[i], audio_frames[i * 2:i * 2 + 2]),
+                            (res_frame, index_list[i], audio_frames[i * 2:i * 2 + 2]),
                             0)
-                        index = index + 1
+                        # index = index + 1
 
                     logger.info(f'total batch time: {time.perf_counter() - start_time}')
             else: