meetween/mumospee_libritts
收藏资源简介:
该数据集是LibriTTS语料库的衍生版本,已转换为更大的parquet文件,以在高性能计算集群上优化I/O性能。它保持了LibriTTS的高质量、多说话人、文本到语音对齐特性,包含超过585小时的英语有声读物录音和相应转录,适用于语音合成和TTS任务的大规模训练。数据集采用parquet文件格式,采样率为24 kHz,包含超过2400个独特说话人,男女声音平衡,总共有375,086个音频片段。
This dataset is a derived version of the LibriTTS corpus, converted into larger parquet files for optimized I/O performance on high-performance computing clusters. It maintains the high-quality, multi-speaker, text-to-speech alignment of LibriTTS, with over 585 hours of English audiobook recordings and corresponding transcriptions, ideal for large-scale training in speech synthesis and TTS tasks. The dataset uses parquet file format, has a sampling rate of 24 kHz, includes over 2,400 unique speakers with balanced male and female voices, and contains a total of 375,086 segments.




