huseinzol05/Malay-TTS-Osman

Malay-TTS-Osman数据集包含两个部分：Wiki and News和Parliament。Wiki and News部分包含24000样本率的超清洁音频，叙述者为ms-MY-OsmanNeural，时长约94.5小时，文本来自马来语维基百科和新闻，句子长度介于2至20个单词。Parliament部分同样具有24000样本率的超清洁音频，叙述者相同，时长约133.2小时，文本来自马来西亚马来语议会，句子长度介于2至25个单词。

The Malay-TTS-Osman dataset consists of two subsets: Wiki and News, and Parliament. The Wiki and News subset contains ultra-clean audio with a sampling rate of 24000 Hz, spoken by ms-MY-OsmanNeural, with a total duration of approximately 94.5 hours. Its text is sourced from Malay Wikipedia and news articles, and the sentence length ranges from 2 to 20 words. The Parliament subset also features ultra-clean audio at 24000 Hz sampling rate, using the identical speaker. It has a total duration of around 133.2 hours, with text originating from Malaysian Malay parliamentary proceedings, and the sentence length falls between 2 and 25 words.

原始信息汇总

Malay-TTS-Osman 数据集概述

数据集属性

Wiki and News 部分

• 采样率: 24000 Hz
• 叙述者: ms-MY-OsmanNeural
• 总时长: 约94.5小时
• 文本来源: 马来语维基百科和新闻
• 句子长度: 2至20个单词

Parliament 部分

• 采样率: 24000 Hz
• 叙述者: ms-MY-OsmanNeural
• 总时长: 约133.2小时
• 文本来源: 马来西亚马来语议会
• 句子长度: 2至25个单词

数据集使用指南

Wiki and News 部分

1. 下载文件:

   • populated-text.json
   • tts-malay-osman.tar.gz

2. 获取音频和转录文本的方法: python import json import soundfile as sf

   with open(populated-text.json) as fopen: texts = json.load(fopen)

   index = 0 text = texts[index] y, sr = sf.read(fmale/{index}.wav)

Parliament 部分

1. 下载文件:

   • populated-parliament.json
   • tts-malay-osman-parliament.tar.gz

2. 获取音频和转录文本的方法: python import json import soundfile as sf

   with open(populated-parliament.json) as fopen: texts = json.load(fopen)

   index = 0 text = texts[index] y, sr = sf.read(fmale-parliament/{index}.wav)