NEWS: This project is now hosted on GitHub This site is being maintained for now, but please move to using GitHub

DynamicSoundEffectInstance is an instance of a sound which plays buffers submitted from the application. This can be used to implement procedural audio generation or to implement streaming. It can be played with 3D positional audio effects, volume and panning control, and pause/resume control. Looping is not supported.

Note that the DynamicSoundEffectInstance does not copy the wave data and instead refers to the data 'owned' by the application. Therefore, the buffers should be kept "live" until all sounds playing from it are finished.

You can use direct low-level XAudio2 interfaces to implement your own streaming or procedural sound submission rather than make use of this class.


#include <Audio.h>


This class supports integer PCM 8-bit or 16-bit data (defaults to 16-bit) with 1 - 8 interleaved channels.

// PCM 44100 Hz, 16-bit, 1 channel
std::unique_ptr<DynamicSoundEffectInstance> effect( new DynamicSoundEffectInstance( audEngine.get(),
        // 'Buffer needed' event handler
    44100, 1 ) )
For exception safety, it is recommended you make use of the C++ RAII pattern and use a std::unique_ptr.

It can optionally support 3D positional audio:

std::unique_ptr<DynamicSoundEffectInstance> effect( new DynamicSoundEffectInstance( audEngine.get(),
        // 'Buffer needed' event handler
    44100, 1, 16,
    SoundEffectInstance_Use3D ) )

Or use 3D positional audio with reverb effects (if AudioEngine was created using AudioEngine_EnvironmentalReverb | AudioEngine_ReverbUseFilters):

std::unique_ptr<DynamicSoundEffectInstance> effect( new DynamicSoundEffectInstance( audEngine.get(),
        // 'Buffer needed' event handler
    44100, 1, 16,
    SoundEffectInstance_Use3D | SoundEffectInstance_ReverbUseFilters ) )

Instance flags

This is a combination of SoundEffectInstance_Use3D and/or SoundEffectInstance_ReverbUseFilters. It defaults to SoundEffectInstance_Default.

There is also a SoundEffectInstance_UseRedirectLFE which is used internally by the library.

Note: SOUND_EFFECT_INSTANCE_FLAGS is used as a typed flag enum. Only operator| is overloaded to combine them, so operations like |= are not available without additional static_cast<> statements.

Playback control

  • Play ()
Starts the playback of the sound. If paused, it resumes playback.
  • Stop ( bool immediate = true )
Stops the playback of the voice. If immediate is true, the sound is immediately halted. Otherwise any 'tails' are played.
  • Resume ()
Resumes playback if the sound is PAUSED.
  • Pause ()
Pauses the sound playback. Note that for a 'game' pause, you should use AudioEngine::Suspend() / Resume() instead of 'pausing' the sounds individually.

Audio data submission

Unlike SoundEffectInstance which obtains the audio data from a SoundEffect or a WaveBank, this class only plays data provided directly to the class via SubmitBuffer.

std::vector<uint8_t> audioBytes;
audioBytes.resize( 44100 * 2 );

GenerateSineWave( reinterpret_cast<int16_t*>( &audioBytes.front() ), 44100, 440 );
std::unique_ptr<DynamicSoundEffectInstance> effect( new DynamicSoundEffectInstance( audEngine.get(),
    [&audioBytes](DynamicSoundEffectInstance* effect)
        int count = effect->PendingBufferCount();

        while( count < 3 )
            effect->SubmitBuffer( &audioBytes.front(), audioBytes.size() );
    }, 44100, 1, 16 ) );



// The 'Buffer needed' event function is called during some later call to Update() as needed

The 'Buffer needed' callback function is invoked whenever there are <= 2 buffers pending or whenever a buffer completes playback. You should submit sufficient data to avoid starving the voice.

This example uses a simple helper routine which fills a buffer with 1 second of a pure sine wave at a given frequency:

void GenerateSineWave( _Out_writes_(sampleRate) int16_t* data, int sampleRate, int frequency )
    const double timeStep = 1.0 / double(sampleRate);
    const double freq = double(frequency);

    int16_t* ptr = data;
    double time = 0.0;
    for( int j = 0; j < sampleRate; ++j, ++ptr )
        double angle = ( 2.0 * XM_PI * freq ) * time;
        double factor = 0.5 * ( sin(angle) + 1.0 );
        *ptr = int16_t( 32768 * factor );
        time += timeStep;

Note: This sample is somewhat contrived and is only for illustrative purposes. If the audio data is static, it's more efficient to use a SoundEffect and then create a SoundEffectInstance from it to play it as a looped sound. Dynamic cases will have different audio data in the buffer on each call to SubmitBuffer. This requires some additional complexity to ensure that each buffer is freed or reused only after it has completed playing.

Volume and panning

  • SetVolume ( float volume )
Sets playback volume. Playback defaults to 1
  • SetPitch ( float pitch )
Sets a pitch-shift factor. Ranges from -1 to +1, playback defaults to 0 (which is no pitch-shifting). This will trigger a C++ exception if the object was created with SoundEffectInstance_NoSetPitch.
  • SetPan ( float pan )
Sets a pan settings: -1 is fully left, +1 is fully right, and 0 is balanced.
Note panning is only supported for mono and stereo sources, and will overwrite any Apply3D settings.

Positional 3D audio

DirectXTK for Audio uses X3DAudio for positional audio computations. To apply a 3D effect to a sound instance, you call Apply3D with the listener location (i.e. where the player/camera is located) and the emitter (i.e. where the sound source is located in 3D dimensions):

AudioListener listener;
listener.SetPosition( ... );

AudioEmitter emitter;
emitter.SetPosition( ... );

effect->Apply3D( listener, emitter );

Note if the instance was created without SoundEffectInstance_Use3D, then calls to Apply3D will result in a C++ exception being thrown. Apply3D will overwrite any SetPan settings.

See AudioListener, AudioEmitter


  • GetState ()
  • GetPendingBufferCount ()
Returns the number of pending buffers submitted to the instance that are either playing or waiting to play.
  • GetFormat ()
Returns a WAVEFORMATEX that describes the audio format consumed by the sound instance. This is always integer PCM data.

Helper methods

  • GetSampleDuration ( bytes )
Returns duration in samples of a buffer of a given size
  • GetSampleDurationMS ( bytes )
Returns duration in milliseconds of a buffer of a given size
  • GetSampleSizeInBytes( duration )
Returns size of a buffer for a duration given in milliseconds

Further reading

Last edited Jul 28, 2015 at 8:05 PM by walbourn, version 19