Ryuichi Yamamoto

WIP: Support for Wavenet vocoder

7

- [x] Add script to generate training data for wavenet vocoder - [ ] Training deepvoice3 for wavenet vocoder - [ ] Training WaveNet vocoder - [x] Add a option...

Attention doesn't work well for downsample_step=1 and outputs_per_step=1

3

Noticed while working on https://github.com/r9y9/deepvoice3_pytorch/pull/21. Trained 300k steps, but the model was not generalized well. Need to figure out how we can improve.

``` % julia-master --check-bounds=no (git)-[master] _ _ _ _(_)_ | A fresh approach to technical computing (_) | (_) (_) | Documentation: http://docs.julialang.org _ _ _| |_ __ _ |...

Clean up the separation of a sized value and a C++ type

7

I'm having a trouble with multiple dispatch with nested Cxx templated types (e.g. `StdVector{StdVector{T}}`). A minimum illustration of the problem is as follows: ``` jl julia> import CxxStd: StdVector julia>...

Tacotron2 TTS demo: Adapt to the latest tensorflow

13

It seems there's no clean way to downgrade tensorflow on google colab, so we might need to update https://github.com/r9y9/Tacotron-2 to make it run on the latest tensorflow.

CMU Arctic MuLaw samples are pitch shifted.

1

Pointed out by @rafaelvalle. This is a stupid bug... [arctic_b0534_samples.zip](https://github.com/r9y9/wavenet_vocoder/files/2041633/arctic_b0534_samples.zip)

Document how to build speech synthesis system for new languages

11

All you need is that - Wav files - Full-context labels - HTS-style question file With those all prepared, it should be very straightforward to implement.

https://github.com/r9y9/pylibfreenect2/blob/288e187ec94fd0df30d483f6c11318820e1d1e06/pylibfreenect2/libfreenect2.pyx#L1026

ref: https://github.com/r9y9/pylibfreenect2/issues/32#issuecomment-276427067

https://github.com/r9y9/nnsvs/issues/90#issuecomment-1079940339