PulseExploreJournal ClubResearchersJournals
Instagram
HomeJournal ClubExplore
Synapse
⌘+K
Synapse
October 9, 2025Open Access

SingNet: Towards a Large-Scale, Diverse, and In-the-Wild Singing Voice Dataset

View Full Paper
Ask AI
Bookmark
Share

Authors

YGYicheng GuCWChaoren WangJZJ. S. Zhang

Discussion

Loading...

Member takes

Overview

This dataset enhances singing voice synthesis and conversion, providing 3000 hours of diverse audio data, indicating strong potential for future applications.

Key Points

  • SingNet comprises 3000 hours of audio data from various languages and styles, expanding the scope for singing voice applications.
  • The dataset addresses the major bottleneck in singing voice synthesis and conversion, providing a foundation for advancements in these fields.
  • A data processing pipeline was developed to extract training data from internet sources, ensuring accessibility for researchers.
  • State-of-the-art models were pre-trained on the dataset and are now open-sourced, fostering collaboration and innovation in the singing voice domain.

Cite This Study

Gu et al. (2025) studied this question.

synapsesocial.com/papers/68e8439a9989581a2fd4df53https://doi.org/10.48550/arxiv.2505.09325
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Singing Voice Data Scaling-up: An Introduction to ACE-Opencpop and ACE-KiSing2024 · 9 citations
  2. 2VISinger2+: End-to-End Singing Voice Synthesis Augmented by Self-Supervised Learning Representation2024
  3. 3CtrSVDD: A Benchmark Dataset and Baseline Analysis for Controlled Singing Voice Deepfake Detection2024
  4. 4CtrSVDD: A Benchmark Dataset and Baseline Analysis for Controlled Singing Voice Deepfake Detection2024 · 2 citations
  5. 5DSUSING: Dual Scale U-Nets for Singing Voice Synthesis2024