Towards generalizing deep-audio fake detection networks

Konstantin Gasenzer, Moritz Wolter

Today's generative neural networks allow the creation of high-quality synthetic speech at scale. While we welcome the creative use of this new technology, we must also recognize th…