Steveeeeeeen/yodas-granary-it-neucodec-10s-20s

Granary Italian VoxPopuli NeuCodec是一个用于NeuTTS微调的数据集，基于NeuCodec编码的意大利语语音数据。源数据来自espnet/yodas-granary数据集中的Italian配置，经过流式处理并保存为Parquet分片。数据行包含文本、编码、持续时间、源数据集信息等列，并应用了文本过滤（仅限意大利语，标签为<it>）以及时长（10-20秒）、字符数（8-260字符）、字符每秒速率（3-32字符/秒）等筛选条件，最大训练样本数为500,000。Granary Italian VoxPopuli NeuCodec is a dataset of NeuCodec-tokenized rows for NeuTTS fine-tuning. The source rows were streamed from espnet/yodas-granary / Italian and uploaded as resumable Parquet shards. It includes columns such as text, codes, duration, source metadata, with text filtering enabled for Italian language (tag: <it>), and constraints on duration (10-20 seconds), text characters (8-260 characters), characters per second (3-32 chars/sec), and a maximum of 500,000 training samples.