How does VoiceMod achieve real time audio conversion with such low CPU?

I've taken a look at all the available open source models for voice conversion and my cpu is like 55%, whereas VoiceMod is using just 8-10%.

Anyone have any clue how they may have achieved this?