数据堂—12小时中文女声娱乐主播风格多情感合成库

12小时娱乐主播风格多情感合成库，由中文母语发音人录制。七种情感数据+语气词。语料音素覆盖均衡，专业语音学家参与标注，精准匹配语音合成的研发需求

This 12-hour entertainment host-style multi-emotional speech synthesis corpus was recorded by native Mandarin speakers. It includes seven types of emotional speech data and sentence-final modal particles. The corpus has well-balanced phoneme coverage, and was annotated with the participation of professional phoneticians, which precisely meets the R&D requirements for speech synthesis.

搜集汇总

背景与挑战

背景概述

该数据集是一个12小时的中文女声娱乐主播风格多情感合成库，专为语音合成测试任务设计。它包含七种情感类型和感叹句的录音，由专业声优在录音室环境中以48kHz、24bit的wav格式录制，并提供了音素边界和韵律标注。数据量分布为中性情感不少于1.6小时，感叹句不少于0.4小时，其余六种情感各不少于1.67小时。