HUI-Audio-Corpus-German

HUI-Audio-Corpus-German是由霍夫应用科技大学创建的一个高质量的德语TTS数据集，包含超过326小时的音频片段及其匹配的转录文本。数据集主要从librivox.org收集，并通过精细的加工流程处理，以确保音频与转录文本之间的高质量对齐。该数据集包含五个主要发言者和117个额外发言者的音频，旨在为单发言者和多发言者TTS模型提供多样性。数据集的创建过程中，采用了自动化下载、精确的音频文件与转录文本对齐、音频/文本标准化等技术。该数据集主要应用于德语的文本到语音转换领域，旨在解决现有德语TTS数据集质量不一和数量有限的问题。

HUI-Audio-Corpus-German is a high-quality German text-to-speech (TTS) dataset developed by Hof University of Applied Sciences. It contains over 326 hours of audio clips paired with their corresponding transcriptions. The dataset is primarily sourced from librivox.org, and undergoes a rigorous processing workflow to ensure high-quality alignment between audio content and their matched transcriptions. It includes audio data from 5 primary speakers and 117 additional speakers, with the goal of providing sufficient diversity for both single-speaker and multi-speaker TTS models. During the dataset construction process, technologies including automated downloading, precise alignment between audio files and their transcriptions, as well as audio and text normalization were utilized. This dataset is mainly applied in the German text-to-speech domain, and is designed to address the issues of inconsistent quality and limited quantity of existing German TTS datasets.