02 · Start

Quickstart

Five recipes. Each one runs as written.

Play a file

import sound_rs

sound_rs.play_audio("music.mp3")

play_audio blocks until the track ends and releases the GIL while it does, so other Python threads keep running. It opens the output device for that one call.

Play with controls

import sound_rs

with sound_rs.Player() as player:
    player.play_file("music.flac")   # returns immediately
    player.volume = 0.6
    player.seek(30.0)                # jump to 0:30
    player.pause()
    player.resume()
    player.wait()                    # block until it ends

The player keeps the stream open, so starting the next track costs milliseconds instead of reopening the device. Details in The player.

Play a NumPy array

import numpy as np
import sound_rs

rate = 44100
t = np.arange(rate * 2) / rate
mono = (0.2 * np.sin(2 * np.pi * 440 * t)).astype(np.float32)

sound_rs.play_array(mono, rate)                       # 1D, mono
sound_rs.play_array(np.stack([mono, mono]), rate)     # 2D, (channels, frames)

Inspect a file

sample_rate, channels, duration = sound_rs.get_audio_info("music.m4a")
print(sample_rate, channels, duration)   # 44100 2 213.4

This reads the container header — about 1 ms regardless of file length. It decodes one packet only for AAC, whose container does not record the channel count.

Decode to NumPy

import sound_rs

audio, rate = sound_rs.load("music.flac")   # float32, (frames, channels)
print(audio.shape, rate)

Decode many files

import sound_rs

clips = sound_rs.load_many(["a.flac", "b.flac", "c.flac"])   # all cores, GIL released
for audio, rate in clips:
    print(audio.shape, rate)

Choose a device

import sound_rs

for index, name, is_default in sound_rs.list_devices():
    print(("*" if is_default else " "), index, name)

sound_rs.play_audio("music.mp3", device="MacBook Pro Speakers")

Continue with The player.