From 2452f9f84baecf017d2747ecc357916560bc5fbb Mon Sep 17 00:00:00 2001
From: yumoqing <yumoqing@gmail.com>
Date: Sat, 14 Sep 2024 20:53:38 +0800
Subject: [PATCH] bugfix

---
 rtcllm/vad.py | 29 ++++++++++++++++++++++++++---
 1 file changed, 26 insertions(+), 3 deletions(-)

diff --git a/rtcllm/vad.py b/rtcllm/vad.py
index 415a654..57b26d9 100644
--- a/rtcllm/vad.py
+++ b/rtcllm/vad.py
@@ -8,6 +8,7 @@ from aiortc import MediaStreamTrack
 from aiortc.contrib.media import MediaBlackhole, MediaPlayer, MediaRecorder, MediaRelay
 import webrtcvad
 import wave
+from scipy.io.wavfile import write
 import numpy as np
 from av import AudioLayout, AudioResampler, AudioFrame, AudioFormat
 
@@ -130,6 +131,19 @@ class AudioTrackVad(MediaStreamTrack):
 			duration = f.samples * 1000 / f.sample_rate + duration
 		return duration
 
+	def frames_resample(self, frames, sr=None):
+		fs = []
+		for f in frames:
+			fs1 = self.resample(f, sample_rate=sr)
+			fs += fs1
+		return fs
+
+	def frames_write_wave(self, frames):
+		ndarr = np.frombuffer(b''.join(frames), dtype=np.int16)
+		fn = temp_file(suffix='.wav')
+		write(fn, frames[0].sample_rate, ndarr)
+		return fn
+
 	async def write_wave(self):
 		"""Writes a .wav file.
 
@@ -137,6 +151,9 @@ class AudioTrackVad(MediaStreamTrack):
 		"""
 
 		"""
+		############
+		# Method:1
+		############
 		audio_data = self.to_mono16000_data()
 		path = temp_file(suffix='.wav')
 		# print(f'temp_file={path}')
@@ -152,8 +169,9 @@ class AudioTrackVad(MediaStreamTrack):
 			await self.onvoiceend(path)
 		# print('************over*******')
 		return
-		"""
-
+		############
+		# Method:2
+		############
 		path = temp_file(suffix='.wav')
 		output_container = av.open(path, 'w')
 		out_stream = output_container.add_stream('pcm_s16le', rate=16000, layout='mono')
@@ -164,4 +182,9 @@ class AudioTrackVad(MediaStreamTrack):
 		output_container.mux(out_stream.encode())
 		output_container.close()
 		return path
-
+		"""
+		f1 = self.frames_write_wave(self.voiced_frames)
+		frames = self.frames_resample(self.voiced_frames, sr=16000)
+		fn = self.frames_write_wave(frames)
+		print(f'source wave filename={f1}, mono 16000 wave filename={fn}')
+		return fn