naijavoices/naijavoices-dataset

NaijaVoices数据集包含1,500小时的伊博语、豪萨语和约鲁巴语的真实语音数据，每种语言各500小时，来自超过5,000名不同的说话者。数据集还包括专家精心策划的文本，并确保了女性代表和年龄范围的平衡分布。数据集被分为多个批次，每个批次大约包含167小时的数据。

The NaijaVoices dataset consists of 1,500 hours of authentic speech data (from over 5,000 diverse speakers!) and expert curated text in Igbo, Hausa, and Yoruba. Each language has 500 hours of data, with adequate female representation and balanced age-range distribution (young to old speakers). The dataset is divided into three batches for each language, with each batch containing approximately 167 hours of audio. A compressed version (84GB) is also provided for quicker access. The dataset is licensed under CC BY-NC 4.0, allowing free use for personal and research purposes with proper credit.