STARS unifies lyric alignment, note transcription, vocal technique detection, and global style prediction into one multi-level neural model that matches or beats several single-task baselines.
Adversarially Trained Multi-Singer Sequence-To-Sequence Singing Synthesizer
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
This paper presents a high quality singing synthesizer that is able to model a voice with limited available recordings. Based on the sequence-to-sequence singing model, we design a multi-singer framework to leverage all the existing singing data of different singers. To attenuate the issue of musical score unbalance among singers, we incorporate an adversarial task of singer classification to make encoder output less singer dependent. Furthermore, we apply multiple random window discriminators (MRWDs) on the generated acoustic features to make the network be a GAN. Both objective and subjective evaluations indicate that the proposed synthesizer can generate higher quality singing voice than baseline (4.12 vs 3.53 in MOS). Especially, the articulation of high-pitched vowels is significantly enhanced.
fields
cs.SD 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
STARS: A Unified Framework for Singing Transcription, Alignment, and Refined Style Annotation
STARS unifies lyric alignment, note transcription, vocal technique detection, and global style prediction into one multi-level neural model that matches or beats several single-task baselines.