数据堂—10.4小时日语女声合成库

10.4小时日语女声合成库，由日本母语发音人参与录制，发音纯正，音色甜美，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 10.4-hour Japanese female speech synthesis corpus was recorded by native Japanese speakers, featuring authentic pronunciation and a sweet timbre. The corpus has balanced phoneme coverage, and was annotated by professional phoneticians to accurately meet the R&D needs of speech synthesis.

搜集汇总

背景与挑战

背景概述

这是一个10.4小时的日语女声语音合成数据集，用于测试日语女声合成模型。数据包含新闻和对话录音，采用48kHz、16bit的WAV单声道格式，发音地道、音质甜美，音素覆盖平衡，并由专业语音学家标注，句子准确率不低于95%。