revert to default mic buffer size, add ui spectrum movement simulation

components/mic.py

@@ -41,7 +41,7 @@ def get_recording(self):
 
     def get_chunk(self):
         # return np.frombuffer(self._chunk_buffer, np.float32).flatten()
-        return np.frombuffer(self._recording_buffer[-(self.buffer_size*4)*2:], np.float32).flatten()
+        return np.frombuffer(self._recording_buffer[-(self.buffer_size*4):], np.float32).flatten()
 
     def start_mic(self):
         # self._chunk_buffer = bytes()

components/ui.py

@@ -70,35 +70,36 @@ def copy_text(self):
 
     def update_spectrum_viz(self, user_name, data, time_color_warning=0):
         if user_name == "You":
-            spectrum = np.abs(scipy.fft.fft(data))
-            amplitude = spectrum.mean()
-            max_percentage = 0.85
-            max_radius = min(self.spectrum_widget.winfo_reqwidth(), self.spectrum_widget.winfo_reqheight()) * max_percentage / 2
-            sensitivity_factor = 1500
-            scaled_radius = min(amplitude * sensitivity_factor, max_radius)
-            self.radius = int(0.8 * self.radius + 0.2 * max(scaled_radius, self.min_radius))
-            self.spectrum_widget.delete("all")
-            current_window_width = self.root.winfo_width()
-            current_window_height = self.root.winfo_height()
-            self.center_x, self.center_y = int(int(current_window_width)/2), int(int(current_window_height)/4)
-            oval_coords = (
-                self.center_x - self.radius,
-                self.center_y - self.radius,
-                self.center_x + self.radius,
-                self.center_y + self.radius
-            )
-            if 0 < time_color_warning < 0.5:
-                color = "#FF0000"
-            elif 0.5 < time_color_warning < 1:
-                color = "#B93C3C"
-            elif 1 < time_color_warning < 1.5:
-                color = "#8A4B4B"
-            elif 1.5 < time_color_warning < 2:
-                color = "#584848"
-            else:
-                color = "#FFFFFF"
-            self.spectrum_widget.create_oval(oval_coords, outline=color, width=2, fill=color)
-            self.spectrum_widget.update()
+            for i in range(3):
+                spectrum = np.abs(scipy.fft.fft(data))
+                amplitude = spectrum.mean()
+                max_percentage = 0.85
+                max_radius = min(self.spectrum_widget.winfo_reqwidth(), self.spectrum_widget.winfo_reqheight()) * max_percentage / 2
+                sensitivity_factor = 1000 + (500 * i)
+                scaled_radius = min(amplitude * sensitivity_factor, max_radius)
+                self.radius = int(0.8 * self.radius + 0.2 * max(scaled_radius, self.min_radius))
+                self.spectrum_widget.delete("all")
+                current_window_width = self.root.winfo_width()
+                current_window_height = self.root.winfo_height()
+                self.center_x, self.center_y = int(int(current_window_width)/2), int(int(current_window_height)/4)
+                oval_coords = (
+                    self.center_x - self.radius,
+                    self.center_y - self.radius,
+                    self.center_x + self.radius,
+                    self.center_y + self.radius
+                )
+                if 0 < time_color_warning < 0.5:
+                    color = "#FF0000"
+                elif 0.5 < time_color_warning < 1:
+                    color = "#B93C3C"
+                elif 1 < time_color_warning < 1.5:
+                    color = "#8A4B4B"
+                elif 1.5 < time_color_warning < 2:
+                    color = "#584848"
+                else:
+                    color = "#FFFFFF"
+                self.spectrum_widget.create_oval(oval_coords, outline=color, width=2, fill=color)
+                self.spectrum_widget.update()
         elif user_name == "Aria":
             self.spectrum_widget.delete("all")
             self.spectrum_widget.update()

components/vad.py

@@ -44,16 +44,16 @@ def __init__(self, params=None):
             speech_pad_ms=30
             )        
 
-    def check(self, mic_chunk, mic_buffer_size, mic_samplerate):
+    def check(self, mic_chunk, chunk_time):
         speech_dict = self.vad_iterator(mic_chunk, return_seconds=False)
         if speech_dict is not None:
             if "start" in speech_dict:
                 self.no_voice_sec = 0
             elif "end" in speech_dict:
-                self.no_voice_sec += mic_buffer_size / mic_samplerate
+                self.no_voice_sec += chunk_time
         else:
             if self.no_voice_sec != 0:
-                self.no_voice_sec += mic_buffer_size / mic_samplerate
+                self.no_voice_sec += chunk_time
                 if self.no_voice_sec > self.no_voice_wait_sec:                
                     self.no_voice_sec = 0
                     self.vad_iterator.reset_states()

configs/default.json

@@ -2,7 +2,7 @@
     "Mic": {
       "params": {
         "samplerate": 16000,
-        "buffer_size": 256,
+        "buffer_size": 512,
         "channels": 1,
         "device": "default"
       }

main.py

@@ -52,11 +52,11 @@ def main(ui, config):
             print("\nShutting down...")
             break
         mic_chunk = mic.get_chunk()
-        if len(mic_chunk) == mic.buffer_size*2:
+        if len(mic_chunk) == mic.buffer_size:
             if not (mic_chunk==mic_last_chunk).all():
                 mic_last_chunk = deepcopy(mic_chunk)
                 ui.update_spectrum_viz("You", mic_chunk, time_color_warning=vad.no_voice_wait_sec - vad.no_voice_sec)
-                vad_status = vad.check(mic_chunk, mic.buffer_size*2, mic.samplerate)
+                vad_status = vad.check(mic_chunk, mic.buffer_size / mic.samplerate)
                 if vad_status is None:
                     mic.reset_recording()
                     skip_sleep = True
@@ -94,7 +94,7 @@ def main(ui, config):
                     # logging.info("respond starts in: " + str(vad.no_voice_wait_sec - vad.no_voice_sec))
                     pass
         if not skip_sleep:
-            time.sleep(mic.buffer_size*2 / mic.samplerate)
+            time.sleep(mic.buffer_size / mic.samplerate)
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description="Aria.")