Audio Examples of paper: WGANSing: A Multi-Voice Singing Voice Synthesizer Based on the Wasserstein-GAN

Pritish Chandna, Merlijn Blaauw, Jordi Bonada, Emilia Gómez

Music Technology Group, Universitat Pompeu Fabra, Barcelona

Examples From NUS-48E [1] Validation Set (The singers are non-native English Speakers)

These are examples from the NUS-48E [1] dataset, on which the system was trained and evaluated

Original Vocals, Re-synthesized using WORLD vocoder Vocals Synthesized Using WGANSing Model With L1 Loss Vocals Synthesized Using NPSS [2,3] Model Vocals Synthesized Using WGANSing Model Without L1 Loss Vocals Synthesized Using WGANSing Model With L1 Loss, With Voice Change Vocals Synthesized Using WGANSing Model With L1 Loss, With Voice and Gender Change
Male Singing Voice
Original Vocals, Re-synthesized using WORLD vocoder Vocals Synthesized Using WGANSing Model With L1 Loss Vocals Synthesized Using NPSS [2,3] Model Vocals Synthesized Using WGANSing Model Without L1 Loss Vocals Synthesized Using WGANSing Model With L1 Loss, With Voice Change Vocals Synthesized Using WGANSing Model With L1 Loss, With Voice and Gender Change
Female Singing Voice

[1] Duan, Zhiyan, et al. "The NUS sung and spoken lyrics corpus: A quantitative comparison of singing and speech." 2013 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference. IEEE, 2013.

[2] Blaauw, Merlijn, and Jordi Bonada. "A Neural Parametric Singing Synthesizer Modeling Timbre and Expression from Natural Songs." Applied Sciences 7.12 (2017): 1313.

[3] Blaauw, Merlijn, et al. “Data efficient voice cloning forneural singing synthesis,” in2019 IEEE International Conference onAcoustics, Speech and Signal Processing (ICASSP), 2019.