OpenSinger
收藏资源简介:
OpenSinger是由浙江大学发布的大规模多歌手中文歌唱语音数据集,旨在解决高保真歌唱语音合成中的数据稀缺问题。该数据集包含50小时的专业歌手录音,涵盖41位女性和25位男性的歌唱表演,所有录音均在专业录音棚完成,确保高质量无噪音。OpenSinger不仅规模大,质量高,而且是首个公开的中文多歌手歌唱语音数据集,适用于多种歌唱语音合成研究。数据集的创建过程包括精心挑选歌曲、组织专业歌手录音,并由专业团队进行标注和处理,确保数据的准确性和可用性。该数据集主要用于推动歌唱语音合成技术的发展,特别是在多歌手模型和未见歌手模型的合成质量提升方面。
OpenSinger is a large-scale multi-singer Chinese singing speech dataset released by Zhejiang University, aiming to address the data scarcity issue in high-fidelity singing voice synthesis. This dataset contains 50 hours of recordings from professional singers, covering singing performances by 41 female and 25 male artists, with all recordings conducted in professional recording studios to ensure high-quality and noise-free audio. Not only is OpenSinger large-scale and high-quality, but it is also the first publicly available Chinese multi-singer singing speech dataset, suitable for a variety of singing voice synthesis research. The dataset creation process includes carefully selecting songs, organizing recording sessions with professional singers, as well as annotation and processing by a professional team to ensure the accuracy and usability of the data. This dataset is primarily used to promote the development of singing voice synthesis technology, especially in improving the synthesis quality of multi-singer models and unseen singer models.




