1 paper · 1 filter
Iwona Christop, Tomasz Kuczyński, Marek Kubis
We present a novel benchmark for voice cloning text-to-speech models. The benchmark consists of an evaluation protocol, an open-source library for assessing the performance of voic…