KazakhTTS2

KazakhTTS2是由智能系统与人工智能研究所创建的开放源代码哈萨克语文本到语音合成数据集，包含271小时的高质量转录数据，涵盖了新闻、书籍和维基百科文章等多种主题。数据集由五位专业演讲者（三位女性和两位男性）录制，每位演讲者至少有25小时的转录音频。创建过程中，数据集通过手动分割和音频文本对齐，确保了数据的质量和准确性。该数据集主要用于构建高质量的TTS系统，解决哈萨克语等低资源语言的语音合成问题，同时也支持其他突厥语系语言的研究。

KazakhTTS2 is an open-source Kazakh text-to-speech synthesis dataset developed by the Institute of Intelligent Systems and Artificial Intelligence. It contains 271 hours of high-quality transcribed data covering a wide range of topics including news, books, and Wikipedia articles. The dataset was recorded by five professional speakers: three females and two males, with each speaker contributing at least 25 hours of transcribed audio. During its development, manual segmentation and audio-text alignment were carried out to ensure the quality and accuracy of the dataset. This dataset is primarily intended for building high-quality TTS systems to address the speech synthesis challenges faced by low-resource languages such as Kazakh, and also supports research on other Turkic languages.