Ryuichi Yamamoto

Updated pypi release. This should be fixed now.

回答が遅くなりすみません。テキストからフルコンテキストラベルを生成する方法は、https://r9y9.github.io/pyopenjtalk/notebooks/Demo.html#Run-text-processing-frontend-only を参照してください。

回答が遅くなりすみません。可能です。リポジトリにソースコードが含まれていますので、下記のコードを参考にしていただければと思います。 レシピ： https://github.com/r9y9/ttslearn/blob/ea75a160967316181ea42c2e7534b3c95275ceef/extra_recipes/jvs/multispk_tacotron2_pwg/run.sh#L146-L190 Pythonコード: https://github.com/r9y9/ttslearn/blob/ea75a160967316181ea42c2e7534b3c95275ceef/ttslearn/contrib/multispk_util.py#L156-L177

Any comments are appreciated.

Hi, sorry for the late reply. If I remember correctly, samples in M-AI labs are of low SN ratio, and thus WaveNet might suffer from learning a distribution of clean...

Could you also share the config file(s) for WaveNet? For the generated sample, it seems that the signal gain is too high. I guess there would be a mismatch between...

The harams looks okay. I'd recommend you to double-check acoustic feature normalization differences (if any), and also please check analysis/synthesis quality (not TTS). Pre-emphasis at the data preprocessing stage changes...

As noted in https://github.com/r9y9/wavenet_vocoder/blob/c0ac05e41f9f563421172034e9398633df172b4f/hparams.py#L75, `np.prod(upsample_scales)` must be equal to `hop_size`. This is the reason you got the assertion error. Looks like you are using an old json file. Top-level `upsample_scales`...

Ah, I haven't updated https://github.com/r9y9/wavenet_vocoder/tree/c0ac05e41f9f563421172034e9398633df172b4f/presets, which may confuse you. I will simply delete them.

For pretrained models, please checkout the specific git commit as noted in README.