Singing Voice Conversion with Disentangled Representations of Singer and Vocal Technique Using Variational Autoencoders
Yin-Jyun Luo$^{1, 3}$, Chin-Cheng Hsu$^{2}$, Kat Agres$^{3,4}$, Dorien Herremans$^{1,3}$
Many-to-Many Singer and Vocal Technique Conversion
We demonstrate the audio samples corresponding to Fig. 2 in the paper.
The audio files are all inverted from Mel-spectrograms using Griffin-Lim.
The audio "original Mel-spectrogram" thereby serves as the upper bounds of audio quality.
We first give examples of the six vocal techniques
Notice different singers could express a vocal technique differently, and also sing at different levels of expression over time, which raises ambiguities on defining vocal techniques and poses challenges to data-driven models.
Fig. 2(b) - Vocal Technique Conversion
Addtional Samples for Vocal Technique Conversion
Fig. 2(a) - Singers Conversion
