wx44wx/pipecat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fixing smallwebrtc transport input audio resampling logic.

Browse Source
This commit is contained in:
filipi87
2026-02-11 13:01:25 -03:00
parent 0574167fbd
commit ec2b38dc29
1 changed files with 13 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
src/pipecat/transports/smallwebrtc/transport.py
View File

@@ -233,9 +233,8 @@ class SmallWebRTCClient:
self._out_sample_rate = None self._out_sample_rate = None
self._leave_counter = 0 self._leave_counter = 0
# We are always resampling it for 16000 if the sample_rate that we receive is bigger than that. # Audio resampler - will be configured during setup with target sample rate
# otherwise we face issues with Silero VAD self._audio_in_resampler = None
self._pipecat_resampler = AudioResampler("s16", "mono", 16000)
@self._webrtc_connection.event_handler("connected") @self._webrtc_connection.event_handler("connected")
async def on_connected(connection: SmallWebRTCConnection): async def on_connected(connection: SmallWebRTCConnection):
@@ -375,32 +374,22 @@ class SmallWebRTCClient:
await asyncio.sleep(0.01) await asyncio.sleep(0.01)
continue continue
if frame.sample_rate > self._in_sample_rate: # Resample if needed, otherwise use the frame as-is
resampled_frames = self._pipecat_resampler.resample(frame) frames_to_process = (
for resampled_frame in resampled_frames: self._audio_in_resampler.resample(frame)
# 16-bit PCM bytes if frame.sample_rate != self._in_sample_rate
pcm_array = resampled_frame.to_ndarray().astype(np.int16) else [frame]
pcm_bytes = pcm_array.tobytes() )
del pcm_array # free NumPy array immediately
audio_frame = InputAudioRawFrame( for processed_frame in frames_to_process:
audio=pcm_bytes, # Convert to 16-bit PCM bytes
sample_rate=resampled_frame.sample_rate, pcm_array = processed_frame.to_ndarray().astype(np.int16)
num_channels=self._audio_in_channels,
)
audio_frame.pts = frame.pts
del pcm_bytes # reference kept in audio_frame
yield audio_frame
else:
# 16-bit PCM bytes
pcm_array = frame.to_ndarray().astype(np.int16)
pcm_bytes = pcm_array.tobytes() pcm_bytes = pcm_array.tobytes()
del pcm_array # free NumPy array immediately del pcm_array # free NumPy array immediately
audio_frame = InputAudioRawFrame( audio_frame = InputAudioRawFrame(
audio=pcm_bytes, audio=pcm_bytes,
sample_rate=frame.sample_rate, sample_rate=self._in_sample_rate,
num_channels=self._audio_in_channels, num_channels=self._audio_in_channels,
) )
audio_frame.pts = frame.pts audio_frame.pts = frame.pts
@@ -450,6 +439,7 @@ class SmallWebRTCClient:
self._out_sample_rate = _params.audio_out_sample_rate or frame.audio_out_sample_rate self._out_sample_rate = _params.audio_out_sample_rate or frame.audio_out_sample_rate
self._params = _params self._params = _params
self._leave_counter += 1 self._leave_counter += 1
self._audio_in_resampler = AudioResampler("s16", "mono", self._in_sample_rate)
async def connect(self): async def connect(self):
"""Establish the WebRTC connection.""" """Establish the WebRTC connection."""