Jun 23, 2022 · In this study, we propose a singing voice synthesis model with multi-task learning to use both approaches -- acoustic features for a parametric vocoder and mel ...

Singing voice synthesis (SVS) is a generative model to synthe- size singing voices according to musical scores and lyrics. Al- though the musical scores provide ...

Jun 13, 2024 · The paper presents a well-designed and thorough study on improving singing voice synthesis using a multi-task learning approach and a GAN ...

(2022) Adversarial Multi-Task Learning for Disentangling Timbre and Pitch in Singing Voice Synthesis, Accepted by Interspeech 2022. Demo page. (2021) ...

Adversarial Multi-Task Learning for Disentangling Timbre and Pitch in Singing Voice Synthesis ... Another approach is to predict mel-spectrograms for a neural ...

Adversarial Multi-Task Learning for Disentangling Timbre and Pitch in Singing Voice Synthesis. TW Kim, MS Kang, GH Lee. In Proceedings of Interspeech 2022, 3008 ...

Recent Korean SVS systems [6][7][8] are able to generate mel-spectrograms directly by disentangling the timbre and pitch related spectra without external signal ...

This study defines the identity of the singer with two independent concepts – timbre and singing style – and proposes a multi-singer singing synthesis ...

Adversarial Multi-Task Learning for Disentangling Timbre and Pitch in Singing Voice Synthesis. TW Kim, MS Kang, GH Lee. arXiv preprint arXiv:2206.11558, 2022.

해당 접근 방식은 각각의 feature가 명백하게 구분될 수 있다는 이점을 가지고 있음 · 해당 접근 방식의 한계점은 voice의 quality가 좋지 않다는 것.