A submix voice is used primarily for performance improvements and effects processing.
The IXAudio2SubmixVoice interface has these methods.
Data buffers cannot be submitted directly to submix voices and will not be audible unless submitted to a mastering voice. A submix voice can be used to ensure that a particular set of voice data is converted to the same format and/or to have a particular effect chain processed on the collective result.
IXAudio2SubmixVoice inherits directly from IXAudio2Voice, but does not implement methods specific to submix voices. The interface type exists solely because some of the base class methods are implemented differently for submix voices. Having a separate type for these voices helps client code to distinguish the different voice types and to benefit from C++ type safety.