Conversions of singing samples from the NUS-48E dataset to LJS voice. Upper row, shows the name of source singer.
Middle row, the audio sample to be converted. The bottom row shows the conversion generated by our method.
Conversions of singing samples from the NUS-48E dataset to LCSING voice. Upper row, shows the name of source singer.
Middle row, the audio sample to be converted. The bottom row shows the conversion generated by our method.
Conversions of singing samples from the NUS-48E dataset to VCTK voices. Upper row, shows the name of source singer.
Second row, the audio sample to be converted. Third row shows the conversion generated by our method.
Bottom row, shows the name of the target speaker.
Conversions of between singers from the NUS-48E dataset. Each table presents the conversion of a single sample
to all 12 singers in the dataset (including reconstruction). Upper row, shows the name of target singer.
Bottom row shows the conversion generated by our method.