MaxMax2016

如果训练数据很充足，神经网络能自动学习到发音规则；如果数据较少，转成音标再训练可以减少发音错误；

@jinfagang How did you succeed? I have the error like this : Traceback (most recent call last): File "vits_model.py", line 118, in example_outputs=(vits_output),) File "/usr/local/lib/python3.6/dist-packages/torch/onnx/__init__.py", line 28, in _export result...

@jinfagang my code and model are here：https://github.com/dtx525942103/vits_chinese/issues/3

[vits_Chinese.zip](https://github.com/jaywalnut310/vits/files/7230886/vits_Chinese.zip) I surprise to find VITS has not limit to phoneme length, so amazing

用的DB1那个数据集，它是1万句

@hemath1001 the chinese model :https://github.com/dtx525942103/vits_chinese/issues/3

> > [vits_Chinese.zip](https://github.com/jaywalnut310/vits/files/7230886/vits_Chinese.zip) I surprise to find VITS has not limit to phoneme length, so amazing > > 我很想问一下，后验编码器为啥使用线性谱，不直接使用mel谱呢？我看论文里mel重建损失也是用mel谱计算的。。 论文里面说的，使用线性普的效果比使用mel谱的效果更好

what is the status now? Do you know the training time on LJ? Thanks

@rendchevi good news

every thing is ok, no frequency horizontal line noise