bugfix

rtcllm/vad.py

@@ -27,22 +27,6 @@ def frames_write_wave(frames):
 	output_container.close()
 	return path
 
-def to16000_160_frames(frame):
-	remind_byts = b''
-	frames = resample(frame, sample_rate=16000)
-	ret_frames = []
-	for f in frames:
-		if f.samples == 160:
-			return frames
-	for f in frames:
-		b1 = remind_byts + frame2bytes(f)
-		while len(b1) >= 320:
-			b = b1[:320]
-			b1 = b1[320:]
-			ret_frames.append(bytes2frame(b))
-		remind_byts = b1
-	return ret_frames
-
 def bytes2frame(byts, channels=1, sample_rate=16000):
 	audio_data = np.frombuffer(byts, np.int16)
 	audio_data = audio_data.reshape((channels, -1))
@@ -162,13 +146,28 @@ class AudioTrackVad(MediaStreamTrack):
 	def stop(self):
 		self.running = False
 
+	def to16000_160_frames(frame):
+		frames = resample(frame, sample_rate=16000)
+		ret_frames = []
+		for f in frames:
+			if f.samples == 160:
+				return frames
+		for f in frames:
+			b1 = self.remind_byts + frame2bytes(f)
+			while len(b1) >= 320:
+				b = b1[:320]
+				b1 = b1[320:]
+				ret_frames.append(bytes2frame(b))
+			self.remind_byts = b1
+		return ret_frames
+
 	async def recv(self):
 		frame = await self.track.recv()
 		self.sample_rate = frame.sample_rate
 		duration = (frame.samples * 1000) / frame.sample_rate
 		# print(f'{self.__class__.__name__}.recv(): {duration=}, {frame.samples=}, {frame.sample_rate=}')
 		try:
-			frames = to16000_160_frames(frame)
+			frames = self.to16000_160_frames(frame)
 			for frame in frames:
 				await self.vad.vad_check(frame)
 		except Exception as e: