This repository has been archived by the owner on Nov 11, 2023. It is now read-only.
Replies: 1 comment 1 reply
-
请问你是在哪里寻找底模的呢? |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
使用版本为4.0v1,目前数据集为112个60秒的纯人声数据,基本上都是日语,Epoch到了1729,step为48400。
现阶段推理后除了高音的缺频外并无明显质量问题,但在日语的个别发音中会偶尔出错,比如将「ki」读为「ji」、或「hi」无法正常发音的情况十分多见,但使用同一文件的时候能正确读出的情况也并不少。
因为未能找到合适的底模,所以用了“通用“(似乎仅为中文用,并不包括日语)的进行训练。step在初期10000之前上述情况并不明显,似乎是随着训练的进行越来越严重。
目前能想到的原因:
望得到解答。
Beta Was this translation helpful? Give feedback.
All reactions