Upgraded targeting system: Voice control

To make it more interesting I thought it’d be cool to restrict the machine learning model to a single type of item and add support for voice control to change between types. While I was at it I added the audio from the Portal turrets, because why not?