Introduction to PyAudio

A convenient audio package in Python

Presentation by Jean Cruypenynck/@nonatomiclabs

Why?

  • Audio is fun…
  • but always a problem!

Audio + Computers = 💔 ?

  • huge variety of devices, standards, formats, etc.
  • huge variety of native APIs, often complex

Audio + Python = 💔 ?

  • no “correct” audio module in standard library
  • some attempts: winaudio, Carbon.snd, sunaudiodev, ossaudiodev, …
    • platform-dependent
    • limited features
    • often deprecated

PortAudio

A famous C library to deal with audio

Famous because:

  • cross-platform
  • open-source
  • tried and tested
  • simple interface to deal with audio

PortAudio

PyAudio

  • provides Python bindings for PortAudio
  • allows to play audio
  • allows to record audio
  • open-source
  • cross-platform
import pyaudio
pip install pyaudio

Classes

  1. pyaudio.PyAudio
    • ​start/terminate PortAudio
    • open/close streams
    • access audio devices
  2. pyaudio.Stream
    • ​real audio processing is here!

Basic PyAudio flow

import pyaudio

1. Import the package

2. Create an instance of PyAudio

p = pyaudio.PyAudio()

3. Open an audio stream

stream = p.open(…)

4. Start the stream

stream.start_stream()

5. Close the stream when you're done

stream.close()

Opening the stream

open(rate,
     format,
     channels,
     input=False,
     output=False,
     input_device_index=None,
     output_device_index=None,
     frames_per_buffer=1024,
     start=True,
     input_host_api_specific_stream_info=None,
     output_host_api_specific_stream_info=None,
     stream_callback=None)

What is sound?

open()

parameters​

  • rate: sampling frequency (usually 44.1/48 kHz)
  • format: bit depth (usually 16/24 bits)
  • channels: number of audio channels
  • input: to use the stream as input
  • output: to use the stream as output
  • input_device_index: input device to use
  • output_device_index: output device to use
  • frames_per_buffer: length of the audio buffer
  • start: to start the stream directly after opening
  • input_host_api_specific_stream_info: …
  • output_host_api_specific_stream_info: …
  • stream_callback: function used to process audio
stream_callback

The

def callback(in_data, frame_count,
             time_info, status_flags):
    # PROCESSING
    out_data = in_data
    return (out_data, pyaudio.PaContinue)

Function called every time audio is needed
(full input buffer/empty output buffer)

Example 1

Audio level meter

Initialisation

import sys
import pyaudio
import math
import struct
import time

# Instantiate PyAudio
p = pyaudio.PyAudio()

Stream creation & shutting

# Open stream using callback
stream = p.open(format=pyaudio.paInt16,
                channels=1,
                rate=48000,
                frames_per_buffer=1024,
                input=True,
                output=True,
                stream_callback=callback)

# Close the stream after 10 seconds
time.sleep(10)
stream.close()

Callback definition

def callback(in_data, frame_count, time_info, status):
    levels = []
    for _i in range(1024):
        levels.append(struct.unpack('<h', in_data[_i:_i + 2])[0])
    avg_chunk = sum(levels)/len(levels)
    
    print_audio_level(avg_chunk, time_info['current_time'])
    
    return (in_data, pyaudio.paContinue)

Example 2

Audio player

Initialisation

import pyaudio
import wave
import sys

BUFFER_SIZE = 1024

# Opening audio file as binary data
wf = wave.open(sys.argv[1], 'rb')

# Instantiate PyAudio
p = pyaudio.PyAudio()

Stream creation & shutting

file_sw = wf.getsampwidth()

stream = p.open(format=p.get_format_from_width(file_sw),
                channels=wf.getnchannels(),
                rate=wf.getframerate(),
                output=True)

data = wf.readframes(BUFFER_SIZE)

while data != '':
    stream.write(data)
    data = wf.readframes(BUFFER_SIZE)

stream.stop_stream()
stream.close()

p.terminate()

Other applications

  • audio recorder
  • frequency recognition
  • filtering
  • audio effects
  • …

all real-time!

Thank you for your attention!

nonatomiclabs.com