Python 如何同时录制2个音频源?

Python 如何同时录制2个音频源?,python,pyaudio,python-sounddevice,Python,Pyaudio,Python Sounddevice,我想录制一首音轨并将其保存为2个不同的.wav文件。音频曲目应以约6秒的延迟保存,每个.wav应为12秒长 我试着用多处理和pyaudio来实现它,但我无法让它正常工作 请注意,我是python初学者,这是我关于stackoverflow的第一篇文章 def func1(): #Record and save a 12 seconds long .wav def func2(): #Record and save a 12 seconds long .wav if __name__

我想录制一首音轨并将其保存为2个不同的.wav文件。音频曲目应以约6秒的延迟保存,每个.wav应为12秒长

我试着用多处理和pyaudio来实现它,但我无法让它正常工作

请注意,我是python初学者,这是我关于stackoverflow的第一篇文章

def func1():
  #Record and save a 12 seconds long .wav 
def func2():
  #Record and save a 12 seconds long .wav 
if __name__ == '__main__':
  p1 = Process(target=func1)
  p1.start()
  p2 = Process(target=func2)
  p2.start()
  p1.join()
  p2.join()

#start func2 6 seconds after func1


I would expect a data structure like this:
|---1.wav---|---1.wav---|---1.wav---|
      |---2.wav---|---2.wav---|---2.wav---|
     6sec  12sec 18sec 24sec 30sec 36sec 42sec
编辑: 我想出了一点代码,似乎工作得很好。它有0.144秒的延迟。我很高兴改进了这个代码。此代码使用线程而不是多处理

导入pyaudio
输入波
从线程导入线程
导入时间
从日期时间导入日期时间
格式=pyaudio.paInt16
通道=2
费率=44100
区块=1024
CHUNK1=1024
记录时间=12秒
WAVE\u OUTPUT\u FILENAME1=name=“outputs/OUTPUT\u 1”+datetime.now().strftime(“%m:%d:%Y-”)
WAVE\u OUTPUT\u FILENAME2=name=“outputs/OUTPUT\u 2”+datetime.now().strftime(“%m:%d:%Y-”)
def func1():
而1==1:
全局格式
全球频道
全球汇率
全局块
全球记录(秒)
全局波形输出文件名1
WAVE_OUTPUT_FILENAME1=name=“outputs/output1”#+datetime.now().strftime(“%m:%d:%Y-”)
audio=pyaudio.pyaudio()
流=音频。打开(格式=格式,频道=频道,
速率=速率,输入=真,
帧(每帧缓冲区=块)
打印(“录制…”)
帧=[]
WAVE_OUTPUT_FILENAME1=WAVE_OUTPUT_FILENAME1+datetime.now().strftime(“%H;%M;%S.%f--”)
对于范围内的i(0,int(速率/块*记录秒)):
data=stream.read(块)
frames.append(数据)
WAVE_OUTPUT_FILENAME1=WAVE_OUTPUT_FILENAME1+datetime.now().strftime(“%H;%M;%S.%f”)+“.wav”
打印(“已完成录制”)
#停止录音
stream.stop_stream()
stream.close()
audio.terminate()
波形文件=wave.open(波形输出文件名1,'wb')
波形文件设置通道(通道)
波形文件.设置采样宽度(音频.获取样本大小(格式))
波形文件设置帧率(速率)
波形文件.写格式(b''。连接(帧))
波形文件关闭()
def func2():
时间。睡眠(6)
而1==1:
全局格式
全球频道
全球汇率
全局块1
全球记录(秒)
全局波形输出文件名2
WAVE_OUTPUT_FILENAME2=name=“outputs/output2”#+datetime.now().strftime(“%m:%d:%Y-”)
audio=pyaudio.pyaudio()
流=音频。打开(格式=格式,频道=频道,
速率=速率,输入=真,
帧每缓冲区=块1)
打印(“录制…”)
帧=[]
WAVE_OUTPUT_FILENAME2=WAVE_OUTPUT_FILENAME2+datetime.now().strftime(“%H;%M;%S.%f--”)
对于范围内的i(0,int(速率/块1*记录秒)):
data=stream.read(CHUNK1)
frames.append(数据)
WAVE_OUTPUT_FILENAME2=WAVE_OUTPUT_FILENAME2+datetime.now().strftime(“%H;%M;%S.%f”)+“.wav”
打印(“已完成录制”)
#停止录音
stream.stop_stream()
stream.close()
audio.terminate()
波形文件=wave.open(波形输出文件名2,'wb')
波形文件设置通道(通道)
波形文件.设置采样宽度(音频.获取样本大小(格式))
波形文件设置帧率(速率)
波形文件.写格式(b''。连接(帧))
波形文件关闭()
如果uuuu name uuuuuu='\uuuuuuu main\uuuuuuu':
线程(目标=func1).start()
线程(目标=func2).start()

为什么您认为需要
多处理
?我认为这会使事情复杂化

以6秒(或更小)的块/帧记录,并将正确的帧写入每个文件,怎么样

我有点忘乎所以,为此写了一个很好的类:

import pyaudio
import wave
import time


class OverlappedRecorder:
    def __init__(
        self, secs_per_file, secs_between_file, *,
        num_channels=2, sample_rate=48000,
        sample_format=pyaudio.paInt16,
    ):
        # various constants needed later
        self.num_channels = num_channels
        self.sample_width = pyaudio.get_sample_size(sample_format)
        self.sample_rate = sample_rate
        self.frames_between_start = int(secs_between_file * sample_rate)
        self.frames_per_file = int(secs_per_file * sample_rate)

        # mutable state needed to keep everything going
        self.files = []
        self.frames_till_next_file = 0

        self.pa = pyaudio.PyAudio()
        self.stream = self.pa.open(
            format=sample_format, channels=num_channels,
            rate=sample_rate, frames_per_buffer=1024,
            input=True, start=False,
            stream_callback=self._callback,
        )

    def sleep_while_active(self):
        while self.stream.is_active():
            time.sleep(0.2)

    def begin_wave_file(self):
        "internal function to start a new WAV file"
        path = time.strftime(
            'recording-%Y-%m-%d-%H.%M.%S.wav',
            time.localtime()
        )
        file = wave.open(path, 'wb')
        file.setnchannels(self.num_channels)
        file.setsampwidth(self.sample_width)
        file.setframerate(self.sample_rate)
        self.files.append(file)

    # context manager stuff, recording starts when entered using "with"
    def __enter__(self):
        self.stream.start_stream()
        return self

    # exiting shuts everything down
    def __exit__(self, exc_type, exc_val, exc_tb):
        self.stream.stop_stream()
        self.stream.close()
        self.pa.terminate()

        for file in self.files:
            file.close()

    # called by pyaudio when a new set of frames are ready
    def _callback(self, data, frame_count, time_info, status):
        self.frames_till_next_file -= frame_count
        # see if we need to start a new file
        if self.frames_till_next_file < 0:
            self.frames_till_next_file += self.frames_between_start
            self.begin_wave_file()

        # can't remove from lists while iterating
        # keep a list of files to close and remove later
        done = []
        for file in self.files:
            remain = self.frames_per_file - file.getnframes()

            # add appropriate amount of data to all open files
            if frame_count < remain:
                file.writeframesraw(data)
            else:
                remain *= self.sample_width * self.num_channels
                file.writeframesraw(data[:remain])
                done.append(file)

        # close anything that finished
        for file in done:
            file.close()
            self.files.remove(file)

        # tell pyaudio to keep going
        return (None, pyaudio.paContinue)
将让它运行30秒,或者您可以执行以下操作:

with OverlappedRecorder(12, 6) as rec:
    rec.sleep_while_active()
让它一直运行,直到按Ctrl+C终止程序,或者您可以在其中调用
input()
,使它在按enter键时停止,或者其他任何您喜欢的操作

关于您发布的代码的一些评论:

  • 如果要修改变量,只需声明
    global
    变量
  • 为什么你有独立的功能?为什么不只使用一个函数,然后延迟第二个
    线程
  • 为什么要多次设置
    WAVE\u OUTPUT\u FILENAME1
    ?只需保存
    开始时间
    结束时间
    ,然后一次性格式化字符串
  • 您不必成批地
    read()
    ,如果您知道它将适合内存,只需一次读取所有内容即可
  • 您不需要一直启动和停止录制,只需在每个线程中打开它一次,如果幸运的话,在将wav文件写入磁盘时,样本将积累在缓冲区中
比如:

import pyaudio
import wave
import time
from datetime import datetime
from threading import Thread

FORMAT = pyaudio.paInt16
CHANNELS = 2
RATE = 44100
RECORD_SECONDS = 12


def recorder(prefix):
    audio = pyaudio.PyAudio()
    stream = audio.open(
        format=FORMAT, channels=CHANNELS,
        rate=RATE, input=True,
    )
    try:
        while True:
            start_time = datetime.now()
            print("recording started", start_time)

            data = stream.read(RATE * RECORD_SECONDS, False)

            end_time = datetime.now()
            print("finished", end_time)

            name = f'{prefix}{start_time:%Y-%m-%d-%H-%M-%S.%f}-{end_time:%H-%M-%S.%f}.wav'
            print("writing", name)
            with wave.open(name, 'wb') as waveFile:
                waveFile.setnchannels(CHANNELS)
                waveFile.setsampwidth(audio.get_sample_size(FORMAT))
                waveFile.setframerate(RATE)
                waveFile.writeframes(data)
    finally:
        stream.stop_stream()
        stream.close()
        audio.terminate()


if __name__ == '__main__':
    Thread(target=recorder, args=('outputs/output_1-',)).start()
    time.sleep(6)
    Thread(target=recorder, args=('outputs/output_2-',)).start()
一些不同之处:

  • 使用线程的版本代码要少得多
  • 我的版本允许任意数量的文件,而无需为每个文件使用多个操作系统线程(有Python线程,
    pyaudio
    有一个内部线程负责音频缓冲区)
  • 我的版本保存部分文件

希望所有的帮助都有意义

我不知道这在pyaudio中是否可行。至少对我来说不是。谢谢你所有的工作@Sam Mason。我会试着在接下来的几天再看一遍!我有一个关于代码的一部分及其功能的问题:
done=[]对于self.files中的文件:reserve=self.frames\u per\u file-file.getnframes()#如果frame\u countimport pyaudio
import wave
import time
from datetime import datetime
from threading import Thread

FORMAT = pyaudio.paInt16
CHANNELS = 2
RATE = 44100
RECORD_SECONDS = 12


def recorder(prefix):
    audio = pyaudio.PyAudio()
    stream = audio.open(
        format=FORMAT, channels=CHANNELS,
        rate=RATE, input=True,
    )
    try:
        while True:
            start_time = datetime.now()
            print("recording started", start_time)

            data = stream.read(RATE * RECORD_SECONDS, False)

            end_time = datetime.now()
            print("finished", end_time)

            name = f'{prefix}{start_time:%Y-%m-%d-%H-%M-%S.%f}-{end_time:%H-%M-%S.%f}.wav'
            print("writing", name)
            with wave.open(name, 'wb') as waveFile:
                waveFile.setnchannels(CHANNELS)
                waveFile.setsampwidth(audio.get_sample_size(FORMAT))
                waveFile.setframerate(RATE)
                waveFile.writeframes(data)
    finally:
        stream.stop_stream()
        stream.close()
        audio.terminate()


if __name__ == '__main__':
    Thread(target=recorder, args=('outputs/output_1-',)).start()
    time.sleep(6)
    Thread(target=recorder, args=('outputs/output_2-',)).start()