IndicTTS23
收藏资源简介:
IndicTTS23数据集是由印度理工学院马德拉斯分校创建的,旨在为22种印度语言提供高质量的文本到语音合成数据。该数据集包含880小时的录音数据,目前已收集765小时,涵盖了男性和女性的专业录音,以及相应的英语录音。数据集的创建过程包括文本收集、语音录制和质量控制,确保数据的纯净性和准确性。该数据集主要用于训练高质量的TTS系统,特别是在印度语言的TTS合成中,旨在解决现有数据集质量不高的问题。
IndicTTS23 was developed by the Indian Institute of Technology Madras, with the goal of supplying high-quality text-to-speech (TTS) synthesis data for 22 Indian languages. The dataset has a total planned duration of 880 hours of recorded audio, 765 hours of which have been collected to date. It includes professional speech recordings from both male and female speakers, as well as corresponding English recordings. The dataset creation pipeline encompasses three core steps: text collection, speech recording, and quality control, to ensure the purity and accuracy of the data. This dataset is primarily designed for training high-quality TTS systems, particularly for TTS synthesis tasks focused on Indian languages, with the aim of addressing the substandard quality issue of existing relevant datasets.




