How to play base64-encoded sound in Python

Question

I want to play a base64-encoded sound in Python, I've tried using Pygame.mixer but all I get is a hiss of white noise.

This is an example of my code:

import pygame

coinflip = b'data:audio/ogg;base64,T2dnUwACAAAAAAAAAACYZ...'  # Truncated for brevity
flip  = pygame.mixer.Sound(coinflip)
ch = flip.play()
while ch.get_busy():
    pygame.time.wait(100)

The pygame mixer works well if I import a wav/mp3/ogg file, but I want to write a compact self-contained program that doesn't need external files, so I'm trying to embed a base64 encoded version of the sound in the Python code.

NB: The solution doesn't need to be using pygame, but it would be preferable since I'm already using it elsewhere in the program.

Answer 1

The reason you hear white noise is because you try to play audio data with a diffrent encoding then expected.

I think the documentation is not 100% clear about this, but it states that a Sound object represents actual sound sample data. It can be loaded from a file or a buffer. Apparently, when using a buffer, it does expect raw sample data, not some base64-encoded data (and not even raw MP3 or OGG file data). Note that there has been an issue reported about this on the GitHub repository.

So there are two things you can do:

1. Get the raw bytes of your sound (e.g. using pygame.mixer.Sound(filename).get_raw(), or for simple sounds you could create them mathematically) and decode that in base64 format.
2. Wrap the original (MP3/OGG encoded) file data in a BytesIO object, which is a file-like object, so the Sound module will treat it like a file and properly decode it.

Note that in both cases, you still need to base64-decode the data first! The pygame module doesn't automatically do that for you.

Since you want a small file, option 2 is the best. But I'll give examples of both solutions.

Example 1

If you have the raw sample data, you could use that directly as the buffer argument for pygame.mixer.Sound(). Note that the sample data must match the frequency, bit size and number of channels used by the mixer. The following is a small example that plays a 400 Hz sine wave tone.

import base64
import pygame

# The following bytes object consists of 160 signed 8-bit samples,
# which are base64 encoded, When played at 8000 Hz, it results in a
# tone of 400 Hz. The duration of the sound is 0.02 Hz, so it should
# be looped 50 times per second for longer sounds.
base64_encoded_sound_data = b'''
gKfK5vj/+ObKp39YNRkHAQcZNViAp8rm+P/45sqngFg1GQcBBxk1WI
Cnyub4//jmyqeAWDUZBwEHGTVYf6fK5vj/+ObKp39YNRkHAQcZNViA
p8rm+P/45sqngFg1GQcBBxk1WH+nyub4//jmyqd/WDUZBwEHGTVYf6
fK5vj/+ObKp39YNRkHAQcZNViAp8rm+P/45sqnf1g1GQcBBxk1WA==
'''

pygame.mixer.init(frequency=8000, size=8, channels=1, allowedchanges=0)
sound_data = base64.b64decode(base64_encoded_sound_data)
sound = pygame.mixer.Sound(sound_data)
ch = sound.play(loops=50)
while ch.get_busy():
    pygame.time.wait(100)

Example 2

If you want to use a MP3 or OGG file (which is generally much smaller), you could do it like the following example

import base64
import io
import pygame

# Your base64-encoded data here.
# NOTE: Do NOT include the "data:audio/ogg;base64," part.
base64_encoded_sound_file_data = b'T2dnUwACAAAAAAAAAACY...'  # Truncated for brevity

pygame.mixer.init()
sound_file_data = base64.b64decode(base64_encoded_sound_file_data)
assert sound_file_data.startswith(b'OggS')  # just to prove it is an Ogg Vorbis file
sound_file = io.BytesIO(sound_file_data)
# The following line will only work with VALID data. With above example data it will fail.
sound = pygame.mixer.Sound(sound_file)
ch = sound.play()
while ch.get_busy():
    pygame.time.wait(100)

I would have preferred to use real data in this example as well, but the smallest useful Ogg file I could find was 9 kB, which would add about 120 long lines of data, and I don't think that is appropriate for a Stack Overflow answer. But if you replace it with your own data (which is hopefully a valid Ogg audio file), it should work.