Skip to content

Dataset · AaronZ345

Multilingual Vocal Corpus with Scores and Singing Techniques

AaronZ345/GTSinger

Singing audio with various vocal techniques and musical scores in six languages for voice generation.

What's inside

Created by Zhejiang University researchers and presented at NeurIPS 2024 Spotlight, the dataset contains 10K to 100K audio and text records. It includes vocal examples with different performance techniques and realistic musical scores. Six languages are supported: Chinese, English, French, Japanese, Korean, and Spanish.

What you can do with it

  • Train singing voice synthesis systems using notes and text
  • Recognize and classify various vocal techniques
  • Fine-tune audio models for vocal music generation
Tasks
Text → audio, Text → speech
Languages
Chinese, English, French, Japanese, Korean, Spanish, German, Russian and 1 more
Size
10K–100K rows
Downloads this month
44.3K
Likes
19

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("AaronZ345/GTSinger")