Opencpop

Opencpop是一个公开的高质量国语歌声语料库，专为歌声合成（SVS）系统而设计。该语料库由100首独特的国语歌曲组成，由专业女歌手录制。所有音频文件均在专业录音室环境中以44,100 Hz采样率录制，具有录音室品质。
所有歌唱录音均已用语音/音符/音素边界和音高类型进行了语音注释。最终数据集包含 3,756 个话语，总共约5.2 小时。测试集由 5 首随机选择的歌曲组成，并提供基线合成结果。
人声是最美妙的乐器之一。让我们为人类创造可用的歌声合成技术。享受！

Opencpop is a publicly available high-quality Mandarin singing corpus designed exclusively for singing voice synthesis (SVS) systems.
This corpus consists of 100 unique Mandarin songs recorded by professional female vocalists. All audio files were captured in professional studio environments at a sampling rate of 44,100 Hz, with studio-grade audio quality.
All singing recordings have been phonetically annotated with speech, note, and phoneme boundaries as well as pitch types. The finalized dataset comprises 3,756 utterances, with a total duration of approximately 5.2 hours. The test set includes 5 randomly selected songs, alongside baseline synthesis results.
The human voice is one of the most exquisite musical instruments. Let us create usable singing synthesis technology for humanity. Enjoy!

搜集汇总

背景与挑战

背景概述

Opencpop是一个专为歌声合成（SVS）系统设计的公开高质量国语歌声语料库，包含100首由专业女歌手录制的国语歌曲，总时长约5.2小时，具有录音室品质的音频和详细的语音注释。该数据集由西北工业大学、同济大学和上海视觉艺术学院于2022年发布，旨在推动歌声合成技术的发展，并提供测试集和基线结果以供评估。