Skip to content

An efficient, cross-platform Rust crate that mixes together audio from different sources.

Notifications You must be signed in to change notification settings

tuzz/audio_mixer

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

45 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AudioMixer

An efficient, cross-platform Rust crate for mixing together audio from different sources, such as sound files and sine waves. It has a minimal interface (iterators), is easily extended and was written with performance in mind.

I wrote this crate because I initially tried to use rodio and ran into lots of problems with performance and with audio playing in the wrong channels. I haven't published this crate yet but you can install it from the GitHub URL.

How to use

use audio_mixer::AudioMixer;

fn main() {
  let mixer = AudioMixer::for_default_device().unwrap();

  mixer.add(Silence);
  mixer.add(Silence);

  mixer.wait(); // Wait until all sources have finished playing (optional).
}

struct Silence;

impl Iterator for Silence {
  type Item = f32;

  fn next(&mut self) -> Option<Self::Item> {
      Some(0.) // Return your audio sample here, or None when finished.
  }
}

Check out the other examples for programs that actually make sounds.

How to extend

The crate contains a few useful things (see below), but it's also easy to extend it with your own iterators. For example, here's how you would write an iterator to reverse the left/right channels of a stereo source.

struct ReverseStereo<S: Iterator<Item=f32>> {
    stereo_source: S,
    left_sample: Option<f32>,
}

impl<S: Iterator<Item=f32>> ReverseStereo<S> {
    pub fn new(stereo_source: S) -> Self {
        Self { stereo_source, left_sample: None }
    }
}

impl<S: Iterator<Item=f32>> Iterator for ReverseStereo<S> {
    type Item = f32;

    fn next(&mut self) -> Option<f32> {
        if let Some(right_sample) = self.left_sample.take() {
            Some(right_sample)
        } else {
            // Samples are channel-interlaced so this works by stashing the left
            // sample on self and yielding the right one in its place. The
            // iterator then yields the stashed sample on the next call.

            self.left_sample = self.stereo_source.next();
            let right_sample = self.stereo_source.next();

            right_sample
        }
    }
}

See examples/reverse_stereo.rs for a working version of the above code.

Conversions

The crate includes IntoChannels and IntoSampleRate structs to help with channel and sample rate conversions. These are needed, for example, when trying to play a 44100Hz mono source on a 48000Hz stereo output device. Without these, the audio will play in the wrong channels and/or at the wrong speed.

Many of the iterators in the crate use a strategy pattern so that unnecessary processing doesn't take place. For example, if you convert 2 channels into 2 channels, the samples will simply be forwarded on (a "no op"). Therefore, you don't need to check these conditions yourself before deciding whether an iterator is needed.

Ogg and Wav decoding

The crate supports ogg and wav decoding (via the lewton and hound crates). You need to enable the ogg or wav crate features respectively to use the OggDecoder or WavDecoder structs.

It should be easy to implement your own decoders for other formats, provided you can produce an iterator of channel-interlaced samples. There are plenty of crates available that should be able to help.

See examples/ogg_file.rs for an example that combines OggDecoder, IntoChannels and IntoSampleRate.

cargo run --example ogg_file --features ogg

Reusing buffers

If you want to play a sound multiple times, it makes sense to write this sound into a buffer first so you can reuse it, rather than reading it from the file system and decoding it each time. The crate provides a ReusableBuffer struct to help with this. See examples/buffer_reuse.rs.

The ReusableBuffer also allows you to seek forwards and backwards through a sound or start it playing from a specific point. See examples/buffer_seeking.rs.

Dynamic controls

If you want to change some parameter of an iterator while it is being read (e.g. change the pitch or volume) there are DynamicUsize and DynamicFloat structs to help with this. See examples/dynamic_controls.rs.

Low-pass filtering

The crate provides a LowPassFilter which is perhaps the most complicated of the available filters. It requires that you tell it which threshold frequencies and sample rates you want to filter in advance to keep expensive computations out of the time-sensitive audio thread. See examples/low_pass.rs.

Controlling playback

The crate provides PausableAudio and StoppableAudio. These allow audio to be paused temporarily and stopped completely via a dynamic control. Their implementations are iterator-based like everything else.

The StoppableAudio iterator can also be used as an 'observer' to determine when a source has finished playing. See examples/pausing_playback.rs and examples/stopping_playback.rs.

Recording playback

The crate provides audio_mixer.start_recording which allows you to capture the stream of samples sent to the audio device. These samples are provided to a closure along with some metadata such as how long the audio has been playing. See examples/recording_audio.rs

You could use this feature, for example, to record audio into a file or to stream over the network. This feature is really intended to 'listen in' on what's playing on the device rather than as an offline processing pipeline since it works in real-time with an audio device.

Optimizations

The crate provides three iterators that might help you to reduce load:

  • PauseWhenMuted: Pauses while the volume is muted. Resumes when the volume is unmuted.
  • SkipWhenMuted: Pauses while muted. Seeks to its expected position when unmuted.
  • StopWhenMuted: Stops immediately when muted. The iterator fuses and returns None.

These iterators can save unnecessary work in the iterator chain when the audio is muted. For example, they can be used to bypass sample rate conversions and low-pass filtering which are relatively expensive operations. The SkipWhenMuted iterator makes use of the seeking capability of ReusableBuffer. See examples/pausing_when_muted.rs and examples/skipping_when_muted.rs.

Ideas for improvement

  • Fix SkipWhenMuted not handling channels properly
  • Make cpal an optional dependency so that you can mix audio together without also having to play it on a device

License

MIT

About

An efficient, cross-platform Rust crate that mixes together audio from different sources.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages