Firoj112/voxcpm-nepali-data
收藏Hugging Face2026-04-23 更新2026-04-26 收录
下载链接：
https://hf-mirror.com/datasets/Firoj112/voxcpm-nepali-data
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是用于微调VoxCPM的尼泊尔语语音音频数据集。数据集来源于ai4bharat/indicvoices_r，包含8,002个音频片段（约30小时），语言为尼泊尔语（ne），许可证为Apache 2.0。数据集经过重新采样至16kHz单声道，标准化至-18 LUFS，去除静音，并移除了超过15秒的片段。数据集分为训练集（7,201个样本）、验证集（400个样本）和测试集（401个样本）。每个样本包含音频文件、文本、持续时间和ID等信息。
Preprocessed Nepali speech audio for fine-tuning VoxCPM. The dataset is sourced from ai4bharat/indicvoices_r, containing 8,002 audio clips (~30 hours) in Nepali (ne) language, licensed under Apache 2.0. The data has been resampled to 16kHz mono, normalized to -18 LUFS, silence trimmed, and clips longer than 15s removed. The dataset is split into train (7,201 samples), validation (400 samples), and test (401 samples) sets. Each sample includes audio file, text, duration, and ID.
提供机构：
Firoj112
