increase min silence duration for better interrupt handling

jhakulin · Nov 2, 2024 · 0b4f6fa · 0b4f6fa
1 parent 1857fd9
commit 0b4f6fa
Show file tree

Hide file tree

Showing 3 changed files with 14 additions and 13 deletions.
diff --git a/samples/async/main.py b/samples/async/main.py
@@ -76,10 +76,10 @@ def on_speech_start(self):
             asyncio.run_coroutine_threadsafe(self.client.cancel_response(), self.event_loop)
             self.cancelled = True
 
-            current_item_id = self.event_handler.get_current_conversation_item_id()
-            current_audio_content_index = self.event_handler.get_current_audio_content_id()
-            logger.info(f"Truncate the current audio, current item ID: {current_item_id}, current audio content index: {current_audio_content_index}")
-            asyncio.run_coroutine_threadsafe(self.client.truncate_response(item_id=current_item_id, content_index=current_audio_content_index, audio_end_ms=1000), self.event_loop)
+            #current_item_id = self.event_handler.get_current_conversation_item_id()
+            #current_audio_content_index = self.event_handler.get_current_audio_content_id()
+            #logger.info(f"Truncate the current audio, current item ID: {current_item_id}, current audio content index: {current_audio_content_index}")
+            #asyncio.run_coroutine_threadsafe(self.client.truncate_response(item_id=current_item_id, content_index=current_audio_content_index, audio_end_ms=1000), self.event_loop)
 
             # Restart the audio player
             self.event_handler.audio_player.drain_and_restart()
@@ -295,7 +295,8 @@ async def main():
             tools=functions.definitions,
             tool_choice="auto",
             temperature=0.8,
-            max_output_tokens=None
+            max_output_tokens=None,
+            voice="ballad",
         )
 
         # Define AudioStreamOptions
@@ -336,7 +337,7 @@ async def main():
                 "window_duration": 1.0,
                 "silence_ratio": 1.5,
                 "min_speech_duration": 0.3,
-                "min_silence_duration": 0.3
+                "min_silence_duration": 1.0
             },
             enable_wave_capture=False
         )

diff --git a/samples/main.py b/samples/main.py
@@ -67,10 +67,10 @@ def on_speech_start(self):
             logger.info("Cancelling response.")
             self._client.cancel_response()
 
-            current_item_id = self._event_handler.get_current_conversation_item_id()
-            current_audio_content_index = self._event_handler.get_current_audio_content_id()
-            logger.info(f"Truncate the current audio, current item ID: {current_item_id}, current audio content index: {current_audio_content_index}")
-            self._client.truncate_response(item_id=current_item_id, content_index=current_audio_content_index, audio_end_ms=1000)
+            #current_item_id = self._event_handler.get_current_conversation_item_id()
+            #current_audio_content_index = self._event_handler.get_current_audio_content_id()
+            #logger.info(f"Truncate the current audio, current item ID: {current_item_id}, current audio content index: {current_audio_content_index}")
+            #self._client.truncate_response(item_id=current_item_id, content_index=current_audio_content_index, audio_end_ms=1000)
 
             # Restart the audio player
             self._event_handler.audio_player.drain_and_restart()
@@ -271,7 +271,7 @@ def main():
             tool_choice="auto",
             temperature=0.8,
             max_output_tokens=None,
-            voice="echo",
+            voice="ballad",
         )
 
         # Define AudioStreamOptions
@@ -309,7 +309,7 @@ def main():
                 "window_duration": 1.0,
                 "silence_ratio": 1.5,
                 "min_speech_duration": 0.3,
-                "min_silence_duration": 0.3
+                "min_silence_duration": 1.0
             },
             enable_wave_capture=False
         )

diff --git a/samples/utils/vad.py b/samples/utils/vad.py
@@ -6,7 +6,7 @@
 
 class VoiceActivityDetector:
     def __init__(self, sample_rate, chunk_size, window_duration=1.0,
-                 silence_ratio=1.5, min_speech_duration=0.3, min_silence_duration=0.3):
+                 silence_ratio=1.5, min_speech_duration=0.3, min_silence_duration=1.0):
         """
         Initialize the Voice Activity Detector (VAD).