[ADD]add logic of loop frame

human/audio_inference_handler.py

@@ -31,6 +31,8 @@ class AudioInferenceHandler(AudioHandler):
 
         self._is_running = True
         self.last_direction = 1
+        self.startfrom = 0
+        self.frame_indexes = []
         self._exit_event = Event()
         self._run_thread = Thread(target=self.__on_run, name="AudioInferenceHandlerThread")
         self._exit_event.set()
@@ -103,9 +105,10 @@ class AudioInferenceHandler(AudioHandler):
                     print('AudioInferenceHandler not running')
                     break
 
+                self.startfrom = self.frame_indexes[-1] if self.frame_indexes else 0
                 if is_all_silence:
-                    frame_indexes, self.last_direction = human_status.get_index_v2(self.person_config["frame_config"], self.last_direction, batch_size) # [1,3,4]
-                    for i, frame_idx in zip(range(batch_size), frame_indexes):
+                    self.frame_indexes, self.last_direction = human_status.get_index_v2(self.startfrom, self.person_config["frame_config"], self.last_direction, batch_size) # [1,3,4]
+                    for i, frame_idx in zip(range(batch_size), self.frame_indexes):
                         if not self._is_running:
                             break
                         # self.on_next_handle((None, mirror_index(length, index),
@@ -119,9 +122,9 @@ class AudioInferenceHandler(AudioHandler):
                     img_batch = []
                     index_list = []
                     # for i in range(batch_size):
-                    frame_indexes,self.last_direction = human_status.get_index_v2(self.person_config["frame_config"], self.last_direction, batch_size) # [1,3,4]
+                    self.frame_indexes,self.last_direction = human_status.get_index_v2(self.startfrom, self.person_config["frame_config"], self.last_direction, batch_size) # [1,3,4]
                     # TODO: 推理状态下获取循环帧逻辑
-                    for i, frame_idx in zip(range(len(mel_batch)), frame_indexes):
+                    for i, frame_idx in zip(range(len(mel_batch)), self.frame_indexes):
                         # idx = mirror_index(length, index + i)
                         index_list.append(frame_idx)
                         face = face_list_cycle[frame_idx]

human/huaman_status.py

@@ -52,7 +52,7 @@ class HumanStatus:
         self._current_frame = (self._current_frame + 1) % self._total_frames
         return index
     
-    def get_index_v2(self, frame_config:list, last_direction:int=1, batch_size:int=5):
+    def get_index_v2(self, startfrom, frame_config:list, last_direction:int=1, batch_size:int=5):
         """
         """
         audio_frame_length = batch_size
@@ -68,7 +68,7 @@ class HumanStatus:
             first_speak,# 刚开始讲话 向讲话移动
             last_speak, # 讲话结束 向静默移动
         )
-        startfrom = start_idx_list[-1]
+        # startfrom = start_idx_list[-1]
         # 一次返回一个batch帧号
         return start_idx_list, last_direction