multilingual-tts/open-bible

这是一个多语言音频-文本数据集，涵盖35种语言（包括阿拉伯语、阿萨姆语、孟加拉语、中央库尔德语、恰蒂斯加尔语、奇切瓦语、达夫罗语、德霍卢奥语、埃维语、加莫语、戈法语、古吉拉特语、海地克里奥尔语、豪萨语、希利盖农语、印地语、伊博语、卡纳达语、基库尤语、林加拉语、卢干达语、马拉雅拉姆语、马拉地语、恩德贝莱语、尼泊尔语、奥罗莫语、旁遮普语、绍纳语、斯瓦希里语、泰米尔语、泰卢固语、土耳其语、特威语（阿库阿佩姆和阿桑特）、乌尔都语、越南语和约鲁巴语）。每个样本包含音频文件、对应的文本转录，以及结构化元数据：testament（可能指圣经新旧约）、book（书卷）、chapter（章节）、verse（经文）、音频时长（秒）和说话者ID。数据集仅提供训练集，各语言配置的样本量约28,000至30,000+，总数据规模从约10GB到44GB不等，适用于语音识别、文本到语音或多语言语音处理研究。

This is a multilingual audio-text dataset covering 35 languages (including Arabic Standard, Assamese, Bengali, Central Kurdish, Chhattisgarhi, Chichewa, Dawro, Dholuo, Ewe, Gamo, Gofa, Gujarati, Haitian Creole, Hausa, Hiligaynon, Hindi, Igbo, Kannada, Kikuyu, Lingala, Luganda, Malayalam, Marathi, Ndebele, Nepali, Oromo, Punjabi, Shona, Swahili, Tamil, Telugu, Turkish, Twi (Akuapem and Asante), Urdu, Vietnamese, and Yoruba). Each sample includes an audio file, corresponding text transcription, and structured metadata: testament (likely referring to Biblical Old/New Testament), book, chapter, verse, audio duration in seconds, and speaker ID. The dataset provides only training splits, with approximately 28,000 to 30,000+ samples per language configuration, and total data sizes ranging from about 10GB to 44GB. It is suitable for speech recognition, text-to-speech, or multilingual speech processing research.