After an application has created an audio-source object, its next task is to find a suitable engine and speech-recognition mode, and then create an instance of the engine object to represent the mode. An engine typically provides an assortment of speech-recognition modes that can be used to recognize speech in different languages, dialects, or audio sampling rates. The engine object is supplied by the engine vendor.
An application finds a suitable engine and speech-recognition mode by creating a speech-recognition enumerator and using it to find a mode and engine that meet the application's criteria. An application can also enumerate all available modes and select an engine and mode from the list.