transports: allow adding a vad analyzer to BaseInputTransport

examples/foundational/07-interruptible.py

@@ -18,6 +18,7 @@ from pipecat.processors.aggregators.llm_response import (
 from pipecat.services.elevenlabs import ElevenLabsTTSService
 from pipecat.services.openai import OpenAILLMService
 from pipecat.transports.services.daily import DailyParams, DailyTransport
+from pipecat.vad.silero import SileroVADAnalyzer
 
 from runner import configure
 
@@ -41,6 +42,7 @@ async def main(room_url: str, token):
                 audio_out_enabled=True,
                 transcription_enabled=True,
                 vad_enabled=True,
+                vad_analyzer=SileroVADAnalyzer()
             )
         )
 

src/pipecat/transports/base_transport.py

@@ -6,12 +6,16 @@
 
 from abc import ABC, abstractmethod
 
+from pydantic import ConfigDict
 from pydantic.main import BaseModel
 
 from pipecat.processors.frame_processor import FrameProcessor
+from pipecat.vad.vad_analyzer import VADAnalyzer
 
 
 class TransportParams(BaseModel):
+    model_config = ConfigDict(arbitrary_types_allowed=True)
+
     camera_out_enabled: bool = False
     camera_out_is_live: bool = False
     camera_out_width: int = 1024
@@ -27,6 +31,7 @@ class TransportParams(BaseModel):
     audio_in_channels: int = 1
     vad_enabled: bool = False
     vad_audio_passthrough: bool = False
+    vad_analyzer: VADAnalyzer | None = None
 
 
 class BaseTransport(ABC):

src/pipecat/transports/services/daily.py

@@ -161,12 +161,6 @@ class DailyTransportClient(EventHandler):
             "speaker", sample_rate=self._params.audio_in_sample_rate, channels=self._params.audio_in_channels)
         Daily.select_speaker_device("speaker")
 
-        self._vad_analyzer = None
-        if self._params.vad_enabled:
-            self._vad_analyzer = WebRTCVADAnalyzer(
-                sample_rate=self._params.audio_in_sample_rate,
-                num_channels=self._params.audio_in_channels)
-
     @property
     def participant_id(self) -> str:
         return self._participant_id
@@ -174,12 +168,6 @@ class DailyTransportClient(EventHandler):
     def set_callbacks(self, callbacks: DailyCallbacks):
         self._callbacks = callbacks
 
-    def vad_analyze(self, audio_frames: bytes) -> VADState:
-        state = VADState.QUIET
-        if self._vad_analyzer:
-            state = self._vad_analyzer.analyze_audio(audio_frames)
-        return state
-
     def send_message(self, frame: DailyTransportMessageFrame):
         self._client.send_app_message(frame.message, frame.participant_id)
 
@@ -435,6 +423,12 @@ class DailyInputTransport(BaseInputTransport):
         self._video_renderers = {}
         self._camera_in_queue = queue.Queue()
 
+        self._vad_analyzer = params.vad_analyzer
+        if params.vad_enabled and not params.vad_analyzer:
+            self._vad_analyzer = WebRTCVADAnalyzer(
+                sample_rate=self._params.audio_in_sample_rate,
+                num_channels=self._params.audio_in_channels)
+
     async def start(self, frame: StartFrame):
         if self._running:
             return
@@ -461,7 +455,10 @@ class DailyInputTransport(BaseInputTransport):
         await self._client.cleanup()
 
     def vad_analyze(self, audio_frames: bytes) -> VADState:
-        return self._client.vad_analyze(audio_frames)
+        state = VADState.QUIET
+        if self._vad_analyzer:
+            state = self._vad_analyzer.analyze_audio(audio_frames)
+        return state
 
     def read_raw_audio_frames(self, frame_count: int) -> bytes:
         return self._client.read_raw_audio_frames(frame_count)

src/pipecat/vad/silero.py

@@ -39,11 +39,10 @@ def int2float(sound):
         return sound
 
 
-class SileroVAD(FrameProcessor, VADAnalyzer):
+class SileroVADAnalyzer(VADAnalyzer):
 
-    def __init__(self, sample_rate=16000, audio_passthrough=False):
+    def __init__(self, sample_rate=16000):
-        FrameProcessor.__init__(self)
+        super().__init__(sample_rate=sample_rate, num_channels=1)
-        VADAnalyzer.__init__(self, sample_rate=sample_rate, num_channels=1)
 
         logger.debug("Loading Silero VAD model...")
 
@@ -52,7 +51,6 @@ class SileroVAD(FrameProcessor, VADAnalyzer):
         )
 
         self._processor_vad_state: VADState = VADState.QUIET
-        self._audio_passthrough = audio_passthrough
 
         logger.debug("Loaded Silero VAD")
 
@@ -74,6 +72,15 @@ class SileroVAD(FrameProcessor, VADAnalyzer):
             logger.error(f"Error analyzing audio with Silero VAD: {e}")
             return 0
 
+
+class SileroVAD(FrameProcessor):
+
+    def __init__(self, sample_rate=16000, audio_passthrough=False):
+        super().__init__()
+
+        self._vad_analyzer = SileroVADAnalyzer(sample_rate=sample_rate)
+        self._audio_passthrough = audio_passthrough
+
     #
     # FrameProcessor
     #
@@ -89,7 +96,7 @@ class SileroVAD(FrameProcessor, VADAnalyzer):
     async def _analyze_audio(self, frame: AudioRawFrame):
         # Check VAD and push event if necessary. We just care about changes
         # from QUIET to SPEAKING and vice versa.
-        new_vad_state = self.analyze_audio(frame.audio)
+        new_vad_state = self._vad_analyzer.analyze_audio(frame.audio)
         if new_vad_state != self._processor_vad_state and new_vad_state != VADState.STARTING and new_vad_state != VADState.STOPPING:
             new_frame = None