CML-TTS
收藏资源简介:
CML-TTS是一个多语言文本到语音合成数据集,由联邦大学戈亚斯分校的人工智能卓越中心开发。该数据集基于Multilingual LibriSpeech,包含七种语言的音频书籍,旨在为多语言模型提供新的研究可能性。数据集总时长为3,233.43小时,包含613位说话者,采样率为24kHz,适用于训练TTS模型。创建过程中,数据集通过下载原始音频、文本规范化、音频分割和文本验证等步骤处理,确保数据质量。CML-TTS的应用领域主要集中在多语言TTS模型的研究和开发,以解决不同语言环境下语音合成的需求。
CML-TTS is a multilingual text-to-speech synthesis dataset developed by the Center of Excellence in Artificial Intelligence at the Federal University of Goiás. Built upon Multilingual LibriSpeech, the dataset includes audiobooks spanning seven languages, and is designed to offer novel research possibilities for multilingual models. It has a total duration of 3,233.43 hours, encompasses 613 speakers, and features a sampling rate of 24 kHz, making it suitable for training TTS models. During its curation, the dataset underwent processing steps including raw audio download, text normalization, audio segmentation and text validation to guarantee data quality. The primary application domains of CML-TTS lie in the research and development of multilingual TTS models, aimed at fulfilling the speech synthesis requirements across diverse linguistic contexts.




