search

openvpi / DiffSinger

An advanced singing voice synthesis system with high fidelity, expressiveness, controllability and flexibility based on DiffSinger: Singing Voice Synthesis via Shallow Diffusion Mechanism
Apache License 2.0
2.64k stars 275 forks source link

数据集制作3.2 cell 2 #44

Closed QFDG closed 1 year ago

QFDG commented 1 year ago

QQ截图20230205180142 把beam后的数字调大，从100试到了10000还是报错 log导出能找到出错音频文件名字，改完后重试 每次导出的log的文件名都不一样

yqzhishen commented 1 year ago

请你检查是否是在 Windows 下制作数据集，系统编码为 GBK，并且文件名带有中文。 另外，非必要请勿调大 beam，否则可能导致准度下降。

QFDG commented 1 year ago

au响度匹配的时候导出文件自动带了中文名，当时没注意QAQ