Research Article

A Voice Cloning Method Based on the Improved HiFi-GAN Model

Table 5

Parameters, inference speed, MOS, and PESQ scores of different vocoders.

VocoderMOS (CI)PESQParameters (M)Speed on GPUSpeed on CPU

Ground truth4.56    0.084.48
WaveNet3.97    0.063.35×0.002
WaveGlow3.96    0.073.19×5.26×0.13
HiFi-GAN4.25    0.073.6313.94×70.34×2.42
Improved HiFi-GAN4.38    0.063.744.38×78.67×3.17