Add volume to playback, and better Google ASR resource management

hugvey/speech/google.py

@@ -58,13 +58,19 @@ class GoogleVoiceClient(object):
         self.isRunning.set()
         
     def generator(self):
+        print('start generator')
         while not self.toBeShutdown and self.isRunning.is_set():
             try:
                 # set a timeout, as not to wait infinitely for the buffer when
                 # we actually want to restart 
-                yield self.buffer.get(timeout=.2)
+                yield self.buffer.get(timeout=.3)
             except queue.Empty as e:
                 self.logger.debug('empty mic buffer - restart?')
+                print(self.isRunning.isSet())
+        print('stop generator')
+        self.restart = False # don't trigger double restart
+                
+#                 raise RequireRestart("Restart required (generator)")
             
     def setLanguage(self, language_code):
         if self.language_code == language_code:
@@ -77,20 +83,21 @@ class GoogleVoiceClient(object):
     def run(self):
         self.isRunning.set()
         
+        self.speech_client = speech.SpeechClient()
+        config = types.RecognitionConfig(
+            encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
+            sample_rate_hertz=self.src_rate,
+            language_code=self.language_code)
+        self.streaming_config = types.StreamingRecognitionConfig(
+            config=config,
+            interim_results=True)
+        
         while not self.toBeShutdown:
             try:
                 self.logger.info("wait for Google Voice")
                 self.isRunning.wait()
                 self.logger.info("Starting Google Voice")
                 
-                self.speech_client = speech.SpeechClient()
-                config = types.RecognitionConfig(
-                    encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
-                    sample_rate_hertz=self.src_rate,
-                    language_code=self.language_code)
-                self.streaming_config = types.StreamingRecognitionConfig(
-                    config=config,
-                    interim_results=True)
                 
                 audio_generator = self.generator()
                 requests = (types.StreamingRecognizeRequest(audio_content=content)
@@ -137,6 +144,8 @@ class GoogleVoiceClient(object):
                         self.logger.info(f"Text: {transcript}")
                         self.lastNonFinalTranscript = None
                     
+                    if result.is_final:
+                        print("native final")
                     msg = {
                         "event": "speech",
                         "is_final": result.is_final,
@@ -153,6 +162,7 @@ class GoogleVoiceClient(object):
                         self.logger.warn("Stopping voice loop")
                         break
             except RequireRestart as e:
+                self.restart = False
                 self.logger.warn("Restart Google Voice. Language: {}".format(self.language_code))
             except Exception as e:
                 self.logger.critical(f"Crashed Google Voice: {e}")

hugvey/story.py

@@ -66,6 +66,9 @@ class Message(object):
         msg.setStory(story)
         if 'params' in data:
             msg.params = data['params']
+        if not 'vol' in msg.params:
+            # prevent clipping on some Lyrebird tracks
+            msg.params['vol'] = .8
         return msg
     
     def parseForVariables(self):
@@ -191,7 +194,8 @@ class Reply(object):
         
         now = self.forMessage.story.timer.getElapsed()
         diff = now - u.lastUpdate
-        if diff > 2: # time in seconds to force silence in utterance
+        if diff > 5: # time in seconds to force silence in utterance
+            # useful for eg. 'hello', or 'no'
             self.forMessage.story.logger.warn(
                 f"Set finish time for utterance after {diff}s {u.text}"
                 )