Allow defining whether to insert silence in the output transport.

src/pipecat/transports/base_transport.py

@@ -84,6 +84,8 @@ class TransportParams(BaseModel):
         audio_out_mixer: Audio mixer instance or destination mapping.
         audio_out_destinations: List of audio output destination identifiers.
         audio_out_end_silence_secs: How much silence to send after an EndFrame (0 for no silence).
+        audio_out_insert_silence: Insert silence frames when the audio output queue is empty.
+            When False, the transport will wait for audio data instead of inserting silence.
         audio_in_enabled: Enable audio input streaming.
         audio_in_sample_rate: Input audio sample rate in Hz.
         audio_in_channels: Number of input audio channels.
@@ -144,6 +146,7 @@ class TransportParams(BaseModel):
     audio_out_mixer: Optional[BaseAudioMixer | Mapping[Optional[str], BaseAudioMixer]] = None
     audio_out_destinations: List[str] = Field(default_factory=list)
     audio_out_end_silence_secs: int = 2
+    audio_out_insert_silence: bool = True
     audio_in_enabled: bool = False
     audio_in_sample_rate: Optional[int] = None
     audio_in_channels: int = 1

src/pipecat/transports/smallwebrtc/transport.py

@@ -79,14 +79,17 @@ class RawAudioTrack(AudioStreamTrack):
     supporting queued audio data with proper synchronization.
     """
 
-    def __init__(self, sample_rate):
+    def __init__(self, sample_rate: int, insert_silence: bool = True):
         """Initialize the raw audio track.
 
         Args:
             sample_rate: The audio sample rate in Hz.
+            insert_silence: If True, emit silence when the queue is empty. If False,
+                wait until audio data is available.
         """
         super().__init__()
         self._sample_rate = sample_rate
+        self._insert_silence = insert_silence
         self._samples_per_10ms = sample_rate * 10 // 1000
         self._bytes_per_10ms = self._samples_per_10ms * 2  # 16-bit (2 bytes per sample)
         self._timestamp = 0
@@ -131,12 +134,19 @@ class RawAudioTrack(AudioStreamTrack):
             if wait > 0:
                 await asyncio.sleep(wait)
 
-        if self._chunk_queue:
+        if not self._chunk_queue:
+            if self._insert_silence:
+                chunk = bytes(self._bytes_per_10ms)
+            else:
+                while not self._chunk_queue:
+                    await asyncio.sleep(0.005)
                 chunk, future = self._chunk_queue.popleft()
                 if future and not future.done():
                     future.set_result(True)
         else:
-            chunk = bytes(self._bytes_per_10ms)  # silence
+            chunk, future = self._chunk_queue.popleft()
+            if future and not future.done():
+                future.set_result(True)
 
         # Convert the byte data to an ndarray of int16 samples
         samples = np.frombuffer(chunk, dtype=np.int16)
@@ -484,7 +494,10 @@ class SmallWebRTCClient:
         self._video_input_track = self._webrtc_connection.video_input_track()
         self._screen_video_track = self._webrtc_connection.screen_video_input_track()
         if self._params.audio_out_enabled:
-            self._audio_output_track = RawAudioTrack(sample_rate=self._out_sample_rate)
+            self._audio_output_track = RawAudioTrack(
+                sample_rate=self._out_sample_rate,
+                insert_silence=self._params.audio_out_insert_silence,
+            )
             self._webrtc_connection.replace_audio_track(self._audio_output_track)
 
         if self._params.video_out_enabled: