We could use it with for example vadar sentiment, or detect directly upon the audio with things like pitch.

A suggestion was to first remove background (like nvidias cool filter). RTX Voice.