数据堂—50人中文三风格平均音色合成库

50人中文三风格平均音色合成库，由中文母语发音人录制，每人录制客服类、播音类、故事类三种风格的数据，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。

A Mandarin speech synthesis corpus with averaged timbres across three speaking styles, consisting of 50 native Chinese speakers. Each speaker recorded speech data in three distinct styles: customer service, broadcasting, and storytelling. The corpus features phonetically balanced speech materials, was professionally annotated by phoneticians, and precisely meets the R&D requirements of speech synthesis technologies.

搜集汇总

背景与挑战

背景概述

该数据集是一个50人中文三风格平均音色语音合成库，由25名男性和25名女性录制，包含客服、广播和讲故事三种风格，音素覆盖均衡并经过专业标注，适用于语音合成模型的测试任务。数据格式为48kHz、16bit的WAV音频，版权归数据堂所有，遵循Apache 2.0许可证。