Updated documentation, you don't need to initialize the audio subsystem to do in-place format conversion. (Thanks Simon Hug!)